January 30, 2020 | Utrecht, The Netherlands
Abstracts
Conference 2020“Towards Data Driven Health”
enabling data driven health
Abstractbook Health‐RI conference 2020
# Primary contact Title
1 Aaike van Oord Biobanken.nl, transparency for patients and public 2 Talia Santos SQLite4Radiomics: Automated Feature Extraction Integration with
ConQuest DICOM 3 Trynke de Jong FAIRification of The Lifelines Cohort Study and Biobank 4 Jeroen Beliën The iCRF Generator: Generating interoperable electronic case report
forms using online codebooks 5 Sofie Hansen Call for action: register and test BBMRI-NL’s request portal Podium 6 Erik van Iperen BBMRI’s request portal Podium for samples, data, and images from
Dutch biobanks and data collections 7 Arturo Moncada
Torres Showcasing the Personal Health Train: Federated Learning in Real Time using VANTAGE6
8 Harm Buisman Cancer surveillance 9 Kees Ebben Projecting Patient Data onto Clinical Decision Trees in Oncoguide 10 Brenda Hijmans User demo of a public study in the national cBioPortal instance hosted by
Health-RI 11 Derek de Beurs Applying machine learning on health record data from general
practitioners to predict suicidality 12 Carmen Rubio
Alarcón FAIR data management and stewardship in real practice: capture and integration of translational research data from the PLCRC sub-studies MEDOCC, MEDOCC-CrEATE and PROVENC3
13 Rosemarijn Looije Performance Indicators and the infrastructure for health data 14 A.W. (Sandra) van
den Belt-Dusebout Use of encrypted BSN in record linkage of epidemiological cohorts and biobanks with disease to ensure valid linkages with optimal privacy protection.
15 Tim Hulsen The Ten Commandments of Translational Research Informatics 16 Jan Worst A tranSMART driven clinical diagnostic decision support system on
anemia 17 Rob Hooft Data Stewardship Wizard 18 Rob Hooft Data Desks at the University Medical Centers: Facilitating Access to
Expertise on Data Handling 19 Yaron Caspi Changes in the intracranial volume from early adulthood to the sixth
decade of life 20 Annette Gijsbers The PALGA Portal - Streamlining and professionalizing the request,
delivery and use of pathology data and materials. 21 Valeriu Codreanu Generating CT-scans with 3D Generative Adversarial Networks Using
Supercomputers 22 Alberto Traverso Medical images and AI: the need of a big data revolution 23 Betzabel Cajiao Laboratory variation of molecular testing in a Dutch cohort of metastatic
non-small cell lung cancer patients from 2017 24 Evert-Ben van Veen Herziening Gedragscode gezondheidsonderzoek 25 Ronald van Schijndel Supporting your Research; Tools for Data Management and Processing 26 K. Joeri van der Velde FAIR Genomes: Standardizing a meta-data schema for FAIRifying
personal genome data workflows 27 Marian Beekman BBMRI-Omics: Valuable resource of multi-omics data and analysis tools 28 Jack Broeren Uitdagingen bij het bouwen van een Fair Data station 29 Anne-Charlotte Fauvel EATRIS-Plus - a multi-omic toolbox to support cross omic analysis and
data integration in clinical samples 30 Marcel Koek Fully automatic construction of optimal radiomics workflows 31 Hakim Achterberg Fastr workflow engine for reproducible and managed large-scale
processing 32 Marcel Koek Quantitative Imaging Biomarker Storage and Compute Infrastructure 33 Adriaan Versteeg Streamlining manual tasks in large medical imaging studies 34 Trynke de Jong Linkage of Lifelines and PALGA data: Enhancing multidisciplinary
research 35 Celia van Gelder Towards FAIR Data Steward as profession for the Lifesciences 36 Eric Vermeulen Self-initiated donation to a biobank. Should and could biobanks offer this
option?
Abstractbook Health‐RI conference 2020
37 Fatima El Messlaki CBS Microdata Services 38 Martin Boeckhout Ethics review for non-invasive (nWMO) health research: moving towards
a shared approach 39 Evelien van der
Schaaf Metadata matters
40 Erik Flikkenschld Getting started with trusted FAIR data lakes 41 Merel Wassenaar The real world nature of Prospective Dutch ColoRectal Cancer cohort
(PLCRC) 42 Nathalie Hijmering HOVON Pathology Facility and Biobank: Making the right choices for
workflow and data 43 Petros Kalendralis Public radiomics data collections in an open access Semantic Web
(SPARQL) endpoint 44 Matthijs Sloep The FAIRification of clinical data with modular knowlegde graphs. 45 Stuti Nayak Privacy Sensitive Distributed Analysis of Dementia Cohorts from
Hospitals in The Netherlands 46 H. Pieterman Only one copy 47 Inga Tharun Personal Health Train Coalition 48 Arturo Moncada-
Torres Implementation and Deployment of a Federated Logistic Regression
49 Thomas Rooijakkers Secure Log Rank Test in Survival Analysis on Vertically Partitioned Data using Multi-Party Computation
50 Rick Jansen Characterization of depression symptoms using large scale questionnaire data in the Dutch population: a BBMRI-BIONIC study
51 Louis Ter Meer What is a “digital” patient? An ontological approach. 52 Menno de Vries FAIRification at DNA, RNA and protein level in studying colorectal tumor
progression 53 Paula Jansen The Handbook for Adequate Natural Data Stewardship (HANDS) 54 Rogier de Jong SURF Research Access Management, an authorisation and
authentication service optimised for researchers 55 A.E.C. Schroten Administration of research logistics 56 Fariba Ahmadizar Cardio-metabolic profiling - Association of EFV with increased levels of
circulating lipid metabolites 57 Bob van Wijk Amsterdam UMC expertise center for high performance computing 58 Tessa van der Geest EPTRI - European Paediatric Translational Research Infrastructure: a
bridge towards the future of paediatric medicine 59 Rogier van der Stijl Public-private partnerships in biobanking and biobank-related research 60 Rogier van der Stijl Recommendations for sustainable biobanking 61 Peggy Manders COREON – Committee on Regulation and Research 62 Martin Brandt Galaxy in education using the SURF Research Cloud 63 Tessa van der Geest European Paediatric Translational Research Infrastructure (EPTRI): a
survey to map the expertise of the excellence of developmental pharmacology in pan-European countries
64 Nicolien A. van Vliet Thyroid function and metabolomics: from observational research in BBMRI cohorts to causal inference through Mendelian Randomization
65 Karlijn Groenen Applying the FAIR Data principles to a Rare Disease registry: a case study of the VASCA registry
66 Maxime Bos The metabolic profile of arterial calcification in the multi-cohort BBMRI setting
67 Mark Scheffer e/MTIC - Health Data Portal initiative 68 Janine Felix The Pregnancy And Childhood Epigenetics (PACE) Consortium - A
platform for epigenome-wide association meta-analyses 69 Cees Hof The DANS services for sharing, cataloguing and archiving your health
data 70 Daniele Bizzarri Metabolic risk scores: from metabolome to phenotype and back 71 Purva Kulkarni Towards precision diagnostics: Untargeted metabolomics for the
diagnosis of inborn errors of metabolism in individual patients
Abstractbook Health‐RI conference 2020
Floorplan Jaarbeurs Supernova
Abstractbook Health‐RI conference 2020
Abstract: 1 (Demonstration)
Biobanken.nl, transparency for patients and public Aaike van Oord (1), Tieneke Schaaij‐Visser (2, 3), Huig Schipper (4), Theo Mulder (4), Eric
Vermeulen (5), Ilse Broeders (6), Edmar Weitenberg (3), Susanne Rebers (1), Marjanka Schmidt (1,
2)
(1) ELSI Servicedesk/NKI, (2) BBMRI‐NL, (3) Lygature, (4) Patient and public advisory council (BBMRI‐
NL PPAC), (5) VSOP, (6) Lifelines
To create more transparent and accessible communication towards the public about biobanking
research (with clinical and population biobanks), the website Biobanken.nl has been edited and
reinstated as the place for patients, participants and public to find answers to their questions.
The website has been reinstated in co‐operation with a diverse group of stakeholders, varying from
BBMRI‐NL, large biobanks like PALGA and Lifelines and organizations like VSOP and the Patient and
Public Advisory Council of BBMRI‐NL.
At this Health‐RI conference we publicly announce that the new website Biobanken.nl is open for the
public. Plus that it has a new key feature offering patients and the general public answers to
questions they might have about biobanking and the storage and use of their data and tissue. New
questions result in answers that will be added to the website.
> Useful for biobank researchers and patient organizations
Within the network of Health‐RI the website Biobanken.nl offers a clear and easy accessible medium
for communicating difficult and sometimes sensitive information about privacy, legal rights, research
techniques, impact on patients, ethical boundaries, FAIR data stewardship, patient involvement and
financial organisation of biobanking research.
We invite all biobank research professionals to start using Biobanken.nl for their public
communication, apart from specific dissemination of scientific results of their research. In co‐
operation with media like Kennislink.nl and Dutch hospitals, we will communicate the existence of
Biobanken.nl to a broad audience in 2020.
> Answering questions
Transparency is the goal even though biobanking research and its legal and ethical limitations are
sometimes complicated. If information is not available on the website, website visitors are
encouraged to submit their question, to which an answer will be provided. The expert team of the
ELSI Servicedesk, consisting of experts such as legal specialists and ethicists, is available to provide
expert advice if needed.
Acknowledgements I would like to thank Tieneke Schaaij‐Visser for her support to keep Biobanken.nl the national
platform for transparancy about Biobanking , also after rebuilding the website. Tieneke pursuaded
me to change my initial plan to submit an abstract for a poster and propose a demonstration. Indeed
a better way to reach out to all biobanking professionals about the existence of Biobanken.nl andits
use for them towards the broad public.
Keywords: science communication, transparancy, biobanking, ethical, legal and societal issues,
patient communication
Abstractbook Health‐RI conference 2020
Abstract: 2 (Demonstration)
SQLite4Radiomics: Automated Feature Extraction Integration with
ConQuest DICOM Talia Santos (BSc) (1, 2), Lars van Driel (1, 2), Ivan Zhovannik (MSc) (2, 3), René Monshouwer
(PhD) (2)
(1) Fontys University of Applied Sciences, Eindhoven, The Netherlands, (2) Department of Radiation
Oncology, Radboud Institute for Health Sciences, Radboud University Medical Centre, Nijmegen, The
Netherlands, (3) Department of Radiation Oncology (MAASTRO), GROW – School for Oncology and
Development Biology, Maastricht, The Netherlands
Background: Radiomics stands for quantitative medical image analysis for non‐invasive disease
characterization. Radiomic features can then be fed into machine learning models. These models
could become a powerful asset for prognosis and diagnosis prediction, and treatment selection. But
the process of extracting these features oftentimes in unstructured and time‐consuming requires
certain IT skills. Moreover, radiomics extraction is rarely integrated into clinical imaging systems
(PACS), which is crucial for its clinical translation.
Methods: We developed SQLite4radiomics pipeline to standardize radiomics extraction and integrate
it into Conquest DICOM PACS tool.. . The radiomic feature extraction is performed by means of IBSI‐
compliant open‐source Pyradiomics package. The tool was tested with open anonymized imaging
data from a MAASTRO LUNG‐1 cancer cohort hosted on Health‐RI’s XNAT repository.
Functionality: The tool can perform radiomic feature extraction on stored data based on a selected
extraction strategy and a parameter file. The parameter file defines the settings for Pyradiomics. A
separate configuration file allows the researcher to change the region of interest selection strategy,
among other advanced features. The output of this process is saved to a file that can then be used
for machine learning.
Discussion: The goal of this project was to automate the feature extraction process and reduce time
spent on this aspect of research. It was also important to provide a tool that can be reused and
extended upon. The tool is open‐source and is available on GitHub. The pipeline tool is working with
a command line interface. A version with a web‐based interface was developed for internal use at
the Radboudumc and will be presented during the demonstration.
Acknowledgements Would like to thank all the Klinisch Fysici at the Radiotherapy department at the Radboudumc.
Keywords: PACS; DICOM; radiomics; SQLite; ConQuest DB
Abstractbook Health‐RI conference 2020
Abstract: 3 (Demonstration & Pitch)
FAIRification of The Lifelines Cohort Study and Biobank Trynke R. de Jong (1), Gijs L. Faber (1), Ruud van Vliet (2), Morris A. Swertz (3), Aafje Dotinga (4)
(1) Lifelines Cohort & Biobank, (2) Trivento B.V., (3) Molgenis
The Lifelines Cohort Study and Biobank collects longitudinal health‐related data and samples from
~167,000 inhabitants of the northern parts of the Netherlands (including children and elderly). Our
rapidly expanding collection is available for researchers working in the multidisciplinary field of
healthy ageing. Furthermore, researchers may design additional studies to collect extra biological
samples, physical measurements or questionnaire data from our participants.
In 2019 we completely restructured our database to increase the FAIR‐ness of our data, by i)
streamlining the import of novel data and metadata from various sources into a data platform, ii)
facilitating the release of customized datasets to researchers, and iii) increasing our (meta)data
quality. This resulted in a new data model and a new data platform developed on AWS with our ICT‐
partner Trivento (www.trivento.nl), and a new data catalogue developed by our partner Molgenis
(www.molgenis.org). Both developers allowed Lifelines data managers to configure functionalities
(and access some underlying code).
In our new model, each data point is given a unique place along three axes: the WHO‐axis (“which
participant delivered the data point?”), the WHEN‐axis (“in which context was the data point
collected?”), and the WHAT‐axis (“to which variable does the data point belong?”). The three axes
are implemented as filters in the new catalogue, allowing researchers to compile a compact, pre‐
filtered data order and to determine the number of participants who delivered a given variable in a
given context.
A fourth axis, HOW, represents the protocols used to collect the data (see for more detail our new
metadatawiki: http://wiki‐lifelines.web.rug.nl/doku.php). Combined, the four axes enable the
identification, processing, and communication of quality issues (i.e. data points or data sets that do
not properly adhere to a standard protocol) at various levels. In addition, the model ensures the
rapid incorporation of secondary data developed by our expert users.
Acknowledgements None
Keywords: Lifelines ‐ platform ‐ data model ‐ catalogue ‐ FAIR
Abstractbook Health‐RI conference 2020
Abstract: 4 (Demonstration & Pitch)
The iCRF Generator: Generating interoperable electronic case report
forms using online codebooks Sander de Ridder (1), Jan‐Willem Boiten (2), Gerrit Meijer (3) , Jeroen Beliën (1)
(1) Amsterdam UMC, Vrij Universiteit Amsterdam (2) Lygature, Utrecht, (3) Netherlands Cancer
Institute
Semantic interoperability of clinical data is essential to preserve its meaning and intent when the
data is exchanged, re‐used or integrated with other data. Achieving semantic operability requires the
use of a communication standard, such as HL7, as well as (functional) information standards.
Manually mapping clinical data to a medical thesaurus such as SNOMED CT is complicated and
requires expert knowledge of both the dataset, including its context, and the thesaurus. As an
alternative, the (re‐)use of codebooks, data definitions which may already have been mapped to a
thesaurus, can be a viable approach.
We’ve developed the iCRF Generator, a Java program which can generate the core of an
interoperable electronic case report form (iCRF) for three of the major electronic data capture
systems (EDCs): OpenClinica 3, Castor EDC and REDCap. To build their CRFs, users can select one or
more items from established codebooks, available from an online system called ART‐DECOR. ART‐
DECOR is an open‐source tool suite that supports the creation and maintenance of HL7 templates
and allows the storage of dataset definitions. Nictiz, the centre of expertise for eHealth and the
Dutch SNOMED‐CT release centre, facilitates ART‐DECOR to create health information standards that
are publicly accessible. The iCRF Generator currently provides access to six of these codebooks,
amongst which the Basic Health Data Set (Basisgegevensset Zorg) and the Clinical Building Blocks
(Zorginformatiebouwstenen). By providing an easy to use method to create CRFs for multiple EDCs
based on the same codebooks, interoperability can be more easily attained.
Acknowledgements We thank Jan‐Willem Boiten (Lygature), Gerben Rienk Visser (Trial Data Solutions) and Maarten
Ligtvoet (Nictiz) for reviewing the paper and providing invaluable suggestions. We also thank Wessel
Sloof (UMCG) for testing the generated REDCap exports.
Keywords: Interoperability, eCRF, iCRF, Codebook, FAIR, Software, EDC, Clinical data
Abstractbook Health‐RI conference 2020
Abstract: 5 (Demonstration)
Call for action: register and test BBMRI‐NL’s request portal Podium Erik van Iperen (1), Sofie Hansen (2), Tieneke Schaaij‐Visser (2), David van Enckevort (3), Jeroen
Beliën (4), Morris Swertz (3), Folkert van Kemmendade (5), Jan‐Willem Boiten (2)
(1) Durrer Center for Cardiovascular Research, (2) BBMRI‐NL/Lygature, Utrecht, (3) Department of
Genetics, University Medical Center Groningen,University Groningen,Groningen, The Netherlands, (4)
Amsterdam UMC, Vrije University Amsterdam, department of pathology, Amsterdam, The
Netherlands, (5) ErasmusMC, Rotterdam, The Netherlands
Background/information:
The exchange of samples and data from health registries, health databases, image archives and
biobanks to researchers, is often still administered through e‐mails, fax or telephone. This can make
the management of a request difficult and may hinder the efficient use of valuable resources.
Therefore, within the BBMRI‐NL 2.0 project, we have developed a generic national request portal
‘Podium’.
Methods:
‘Podium’ was developed together with The Hyve, an open‐source software company, and with the
valuable input of various existing request procedures in the Netherlands such as: the BIOS‐
consortium, PSI, PALGA, Lifelines, GO‐NL and PHARMO. Podium is currently in production and can be
used free of cost by researchers and organizations alike for requesting and managing requests of
samples and data in order to facilitate and stimulate efficient, optimal and shared use of available
resources within the Netherlands. Podium supports the following steps (including linked requests): a
generic request form, evaluation and approval of a request, track and trace of the data, and sample
release. It is also possible to link Podium to currently existing back‐end systems, such as has been
done at the NKI with ART.
Results:
In Podium we currently have 74 users and 17 organizations registered, including Go‐NL, BIOS, PSI,
NKI, Pharmo, and the NELSON study. However, we need your input! To enable a national one‐stop
shop request portal, including linked requests ‐ we need more organizations to register. To then
further improve our services, we need your feedback on how Podium works, and we are looking for a
use case to test our first request.
Conclusion:
Based on your input and wishes, we will submit a request for change to update the functionality of
Podium where needed. We hope you will join us in this effort in making existing data more easily
accessible for research.
Acknowledgements BBMRI‐NL
Keywords: FAIR, generic data request tool
Abstractbook Health‐RI conference 2020
Abstract: 6 (Demonstration)
BBMRI’s request portal Podium for samples, data, and images from
Dutch biobanks and data collections Erik van Iperen (1), Sofie Hansen (2), David van Enckevort (2, 3), Jeroen Beliën (4), Morris Swertz
(3), Jan‐Willem Boiten (2)
(1) Durrer Center for Cardiovascular Research, (2) BBMRI‐NL/Lygature, Utrecht, (3) Department of
Genetics, University Medical Center Groningen, University Groningen ,Groningen, The Netherlands,
(4) Amsterdam UMC, Vrije University Amsterdam, department of pathology, Amsterdam, The
Netherlands
The exchange of samples and data from biobank/lab to researcher is traditionally administered
through e‐mails, fax or telephone. Within the BBMRI‐NL 2.0 project, we have developed together
with The Hyve a generic request portal ‘Podium’ for requesting samples and data in order to facilitate
and stimulate efficient, optimal and shared use of available resources within the Netherlands.
Podium is directly linked to the BBMRI‐NL catalogue.
The workflow that is supported in Podium is based on the valuable input of various existing request
procedures in the Netherlands such as: BIOS‐consortium, PSI, PALGA, Lifelines, GO‐NL and PHARMO.
We are currently evaluating the use of Podium at several organization and based on their input and
wishes, we plan to submit a request for change to update the functionality of Podium where needed.
Podium offers all researchers and biobanks a portal supporting the process of requesting samples
and data in a standardized manner, improving quality, reliability and accountability. Podium supports
the following steps: a generic request form, evaluation and approval of a request, track and trace of
the data, and sample release. Every process step is logged. Linked requests are also supported, a
linked request is a request composed of different data types from two or more organizations for
which the resulting datasets need to be linked by subject and/or sample (eg. materials from PALGA
and data from PHARMO).
A link with the catalogue enables users to select interesting samples/data in the catalogue, and
request access to this selection across multiple organizations at once.
The backend system of the NKI, ART, has been successfully linked to Podium using the API.
After evaluating Podium at multiple organizations, we have a list of new features and changes. Next
step will be to prioritize these new features and changes and find additional funding to develop and
implement these.
Acknowledgements NA
Keywords: Request, tool, FAIR, data access
Abstractbook Health‐RI conference 2020
Abstract: 7 (Demonstration)
Showcasing the Personal Health Train:
Federated Learning in Real Time using VANTAGE6 Frank Martin (1), Arturo Moncada‐Torres (1), Melle Sieswerda (1), Johan van Soest (2), and Gijs
Geleijnse (1)
(1) Netherlands Comprehensive Cancer Organization (IKNL), Eindhoven, NL, (2) Maastricht University
Medical Centre+, Maastricht, NL
The growing complexity of cancer diagnosis and treatment needs data sets that are larger and richer
than currently available in a single location or database. This requires incorporating data from
different sources, which is typically done by generating a copy of each dataset and centralizing it by a
trusted party. Unfortunately, sharing patient information is becoming increasingly problematic due
to several risks and challenges, such as loss of data control, logistics of data transmission, and privacy
concerns.
Recently, the Personal Health Train (PHT) has emerged as a platform with the potential to overcome
these limitations. Under this approach, different parties (i.e., the stations) can answer their research
questions (i.e., the trains) collaboratively by exchanging aggregated data and/or statistics while
keeping the underlying data on site, safe and undisclosed. In order to make the PHT a feasible, long‐
term solution it requires a robust, flexible, and reliable infrastructure (i.e., the railways) to handle the
collaborations between parties.
At IKNL, we have developed VANTAGE6, our open‐source priVAcy preserviNg federaTed leArninG
infrastructurE for Secure Insight eXchange. This framework consists mainly of a central server, nodes,
and an interface with the user. We will showcase the capabilities of VANTAGE6 in a real time
Federated Learning demo by computing the average age of a group of participants in a Round‐robin
scenario. Furthermore, we will demonstrate how to perform a privacy‐preserving logistic regression
(as proposed by Li et al., 2015) to predict patient survival (i.e., the train) using the Breast Cancer
Wisconsin Diagnostic Data Set.
Acknowledgements ‐
Keywords: distributed learning, infrastructure, logistic regression
Abstractbook Health‐RI conference 2020
Abstract: 8 (Demonstration)
Cancer surveillance Harm Buisman (1), Guido Out (1)
(1) IKNL
Progress in cancer historically depends on hunches from the field. E.g. a doctor suspects a new
treatment is better, a researcher hypothesizes that centralization of care improves survival or
patients report increased local leukemia rates. These hunches depend on coincidence. What if
nobody gets a hunch? On top of that, doing analyses can be cumbersome and time‐intensive.
Answering research questions requires a researcher to write their own scripts. Also, manually
analyzing multiple cancers, regions, and treatments takes a lot of time. These factors make that the
reduction of the impact of cancer is hampered by chance and time factors.
In the cancer surveillance program at IKNL we develop a cancer monitor with analytics capabilities
based on available data such as the Netherlands Cancer Registry. This allows to 1) automatically
identify surprising findings from the data, 2) provide continuous monitoring on multiple indicators
such as incidence, survival or variation in care, and 3) monitor a variety of dimensions such as tumor
type, region or gender at the same time. The tooling in this monitor provides an additional data
service to inspect found patterns using powerful visualizations and a toolbox of statistical analyses.
With increased identification of promising research directions and a speedup in doing analyses, the
cancer surveillance program at IKNL helps reduce the impact of cancer.
Acknowledgements IKNL for over 30 years of data registered into the Netherlands Cancer Registry
Keywords: cancer, algorithms, data mining, visualization, GIS
Abstractbook Health‐RI conference 2020
Abstract: 9 (Demonstration)
Projecting Patient Data onto Clinical Decision Trees in Oncoguide Kees Ebben, clinical informatician (1), Guido Out, scientific software developer (1), Arturo
Moncada‐Torres, clinical data scientist (1), Thijs van Vegchel, clinical informatician (1)
(1) Netherlands Comprehensive Cancer Organization (IKNL), Eindhoven, NL
In oncology, clinical practice guidelines (CPGs) describe the best practices for treating specific
populations of patients. It has been shown that their implementation improves quality of care by
reducing unwanted variability and bettering outcomes in clinical practice.
CPGs are commonly structured as manuals, where the best practices are described in text through
many chapters. Unfortunately, textual CPGs are frequently ambiguous and inconsistent.
Furthermore, quite often the recommendations for a group of patients are spread through the whole
text, which makes it hard for the reader to get a clear picture of the decision process for a specific
treatment.
To address these issues, we have transformed the textual recommendations into data‐driven clinical
decision trees (CDTs) using FAIR principles. In order to analyze the flow of patients from the trees’
stem (i.e., original pool of patients) to the trees’ leaves (i.e., groups of patients for a specific
recommendation) we projected real‐world patient data for breast and prostate cancer onto the CDTs
in Oncoguide (www.oncoguide.nl). First, we created an inventory of the data‐items of all CDTs. Then,
we performed a delta analysis between these data‐items and the variables available in the
Netherlands Cancer Registry (NCR). We selected the CDTs where all data‐items could be obtained.
Next, we presented the total number of patients that matched the relevant data‐item value, and
showed the number (and percentage) of these patients that were treated according to the CPG. All
occurring treatments were presented and classified as adherent or non‐adherent. Finally, we
visualized this information on top of the CDT at each branch and leaf.
This visualization gives unambiguous and structured insight into guideline adherence. It also has the
potential to aid in the initial development and update of CPGs and could serve retrospectively as
evidence to measure and determine best clinical practice for patient (sub)populations.
Acknowledgements We would like to thank Janneke Verloop, Katja Aben, Aafke Honkoop, Theo de Reijke, and Ignace de
Hingh for their support and fruitful discussions during the development of this project.
Keywords: Clinical Practice Guidelines – Real‐world data – Data Projection – Breast Cancer – Prostate Cancer – Guideline Adherence
Abstractbook Health‐RI conference 2020
Abstract: 10 (Demonstration)
User demo of a public study in the national cBioPortal instance hosted
by Health‐RI N. Cassman (1), B. S. Hijmans (1), M. J. de Vries (1), J. Hudecek (1), M. Bierkens (1), R.J.A Fijneman
(1), R. Azevedo (2), J.‐W. Boiten (2), G.A. Meijer (1)
(1) Department of Pathology, The Netherlands Cancer Institute, Plesmanlaan 121 1066 CX,
Amsterdam, The Netherlands, (2) Lygature, Utrecht, The Netherlands
The cBio Cancer Genomics Portal, ‘cBioPortal’, is an open source data integration platform that
enables (cancer) researchers to view and query complex genomic datasets in a comprehensive
manner. The platform was originally developed by Memorial Sloan Kettering Cancer Center (New
York, USA) (1) and is actively maintained and further developed by an international community. The
original instance of cBioPortal (http://cbioportal.org) currently provides access to data from almost
83000 tumor samples from 273 public studies.
Health‐RI hosts a national instance of cBioPortal in the Netherlands (2), with controlled access. As
operators of the data team of Health‐RI we are involved with safe and secure importing of studies to
the national cBioPortal and would like to show you the possibilities of this platform and how it could
aid you in FAIR data management.
To show cBioPortal’s analysis and visualization capabilities, we would like to invite you to a demo
from the perspective of a researcher wishing to analyze a study’s dataset. We will explore the public
study ‘Low‐Grade Gliomas’ (3) using the national cBioPortal. We will take the viewer through up to
nine steps, as follows:
1. Logging in
2. cBioPortal interface
3. Study view
4. Study exploration
5. Study exploration: Group comparison
6. Study exploration: Patient level view
7. Gene panel‐based view
8. Gene panel‐based view: OncoPrint
9. Gene panel‐based view: Plots
After going through these steps, the viewer will have become familiar with exploring study data in
cBioPortal. The demo will include research questions, encouraging the audience to participate
actively.
Acknowledgements 1. Cerami et al. (2012). The cBio Cancer Genomics Portal: An Open Platform for Exploring
Multidimensional Cancer Genomics Data. Cancer Discovery 2(5): 401–404.
2. https://trait.health‐ri.nl/trait‐tools
3. Johnson et al. (2014) Mutational analysis reveals the origin and therapy‐driven evolution of
recurrent glioma. Science 343(6167):189‐193.
Keywords: cBioPortal, Health‐RI, demo
Abstractbook Health‐RI conference 2020
Abstract: 11 (Poster & Pitch)
Applying machine learning on health record data from general
practitioners to predict suicidality Kasper van Mens (1), Elke Elzinga (2), Mark Nielen (3), Joran Lokkerbol (4), Rune Poortvliet (3), Gé
Donker (3), Marianne Heins (3), Joke Korevaar (3), Michel Dückers (3), Claire Aussems (3), Marco
Helbich (5), Bea Tiemens (6), Renske Gilissen (2), Aartjan Beekman (7), Derek de Beurs (3)
(1) Altrecht Mental Healthcare, Utrecht, The Netherlands, (2) 113 Suicide Prevention, Amsterdam,
The Netherlands, (3) Nivel, Netherlands Institue for Health Services Research, Utrecht, The
Netherlands, (4) Centre of Economic Evaluation & Machine Learning, Trimbos Institutue (Netherlands
Institute of Mental Health), Utrecht, the Netherlands, (5) Human Geography and Spatial Planning,
Utrecht University, Utrecht, The Netherlands, (6) Behavioural Science Institute, Radboud University ,
Nijmegen, The Netherlands, (7) Psychiatry, Amsterdam Public Health (research institute), Amsterdam
UMC, Vrije Universiteit Amsterdam
Background
Suicidal behaviour is difficult to detect in general practice. Machine learning algorithms using
routinely collected data might support General Practitioners (GPs) in the detection of suicidal
behaviour. In this paper, we applied machine learning techniques to support GPs recognize suicidal
behaviour in primary care patients using routinely collected general practice data.
Methods
This case‐control study used data from a national representative primary care database including
over 1.5 million patients (Nivel Primary Care Database). Patients with a suicide (attempt) 2017 were
selected as cases (N = 574) and an at risk control group (N = 207,308) was selected from patients with
psychological vulnerability but without a suicide attempt in 2017. RandomForest was trained on a
small subsample of the data (training set), and externally validated on unseen data (test set).
Results
Almost two‐third (65%) of the cases visited their GP within the last 30 days before the suicide
(attempt). RandomForest showed a positive predictive value (PPV) of 0.05 (0.04 – 0.06), with a
sensitivity of 0.39 (0.32 – 0.47) and area under the curve (AUC) of 0.85 (0.81 – 0.88). Almost all
controls were accurately labelled as controls (specificity = 0.98 (0.97 – 0.98)). Among a sample of 650
at‐risk primary care patients, the algorithm would label 20 patients as high‐risk. Of those, one would
be an actual case and additionally, one case would be missed.
Conclusion
This is the first study to apply machine learning to predict suicidal behaviour using general practice
data. Our results showed that these techniques can be used as a complementary step in the
identification and stratification of patients at risk of suicidal behaviour. The results are encouraging
and provide a first step to use automated screening directly in clinical practice. Additional data from
different social domains, such as employment and education, might improve accuracy.
Acknowledgements Netherlands Organisation for Health Research and Development (ZONMW), Dutch ministry of Health.
Keywords: routine health care data, random forest, suicide preventie
Abstractbook Health‐RI conference 2020
Abstract: 12 (Poster & Pitch)
FAIR data management and stewardship in real practice: capture and
integration of translational research data from the PLCRC sub‐studies
MEDOCC, MEDOCC‐CrEATE and PROVENC3 Lana Meiqari (1), Carmen Rubio Alarcón (1), Dave E.W. van der Kruijssen (2), Suzanna Schraa (2),
Maaike Koelink (2), Olivier Paping (2), Miranda van Dongen (1), Mirthe Lanfermeijer (1), Menno de
Vries (1), Noriko Cassman (1), Brenda Hijmans (1), Rinus Voorham (3), Mariska Bierkens (1), Veerle
M.H. Coupé (4), Miriam Koopman (2), Gerrit A. Meijer (1), Geraldine R. Vink (2, 5), Remond J.A.
Fijneman (1)
(1) The Netherlands Cancer Institute, Amsterdam, The Netherlands, (2) University Medical Center Utrecht,
Utrecht University, Utrecht, The Netherlands, (3) Pathologisch‐Anatomisch Landelijk Geautomatiseerd Archief
(PALGA), Houten, The Netherlands, (4) Amsterdam University Medical Centers, Free University, Amsterdam, The
Netherlands, (5) Netherlands Comprehensive Cancer Organisation, Utrecht, the Netherlands
Background: Colorectal cancer (CRC) is the second most common cancer in the Netherlands. For early
stage disease, there is an unmet clinical need to better define who to treat (or not to treat) with
adjuvant chemotherapy after primary tumor resection. Detection of cell‐free circulating tumor DNA
(ctDNA) in post‐surgery liquid biopsies is a promising biomarker for minimal residual disease (MRD)
and associated with disease recurrence. Therefore, using the Prospective Dutch CRC cohort (PLCRC)
infrastructure, we evaluate the prognostic value of ctDNA post‐surgery in three study cohorts:
MEDOCC (stage II, observational), MEDOCC‐CrEATE (stage II, interventional), and PROVENC3 (stage
III, observational).
Aim: FAIR data management of clinical, biobanking and molecular data from MEDOCC, MEDOCC‐
CrEATE and PROVENC3 to support data collection, analysis and dissemination.
Methods: Following PLCRC informed consent, patients from 25 participating hospitals are registered
SLIM. Clinical data are collected via the Netherlands Cancer Registry (‘NCR’) database and via on‐site
registration in Castor EDC. Tumor tissue and blood samples are collected and shipped to the
Netherlands Cancer institute (NKI). Tissue blocks are requested via the PALGA portal (part of the
Dutch National Tissue Portal project) or sent directly to NKI. Tissue and blood biosample data will be
registered in various NKI systems (Glims, LMS, Molpa). Molecular data will be obtained by targeted
sequencing (Illumina) and analyzed using PGDx bioinformatics pipeline. Clinical and longitudinal
molecular data will be integrated and uploaded to cBioPortal.
Results: An overview of the data flow has been drafted. Collaborations are needed to integrate
clinical data from NCR and Castor and biobanking data. Traceability of biosamples within the NKI
systems is being defined. The data model for cBioPortal is being optimized.
Conclusion: Defining a data‐flow at the start of a project is key to improve translational research
quality and to ensure FAIR data management and Good Clinical Practice.
Acknowledgements
PTRC IT & data project team.
Keywords: FAIR, translational research, PLCRC, ctDNA, NCR, SLIM, PALGA, Castor, cbioPortal.
Abstractbook Health‐RI conference 2020
Abstract: 13 (Poster & Pitch)
Performance Indicators and the infrastructure for health data Rosemarijn Looije (1), Annemarijn Prins‐van Ginkel (1), Astrid Roskes (1)
(1) UMC Utrecht, Department of Business Intelligence, Heidelberglaan 100 Utrecht, The Netherlands
The FAIR (Findable, Accessible, Interoperable and Re‐usable) format provides advantages for all
stakeholders; hospitals, patients, researchers and insurance companies.
At UMC Utrecht we develop management information including strategic dashboards with key
performance indicators, which are evaluated each year and changed according to the focus of the
hospital. In addition, we develop dashboards with a specific focus, for instance, management
information regarding clinical processes and financial outcomes. A healthcare data infrastructure that
follows the FAIR format will provide benefits for the quality of our indicators. Some examples of
these benefits are listed below.
Currently, every (department of a) hospital has its own, non standardized, manner of registering data
and deciding on the business rules that result into an indicator. By making use of the same data in
the same format across hospitals and hospital departments, sharing and comparing will become
easier. This would benefit hospital wide decision‐making.
Second, ambiguous registered data could be disambiguated through the addition of other
information. As the human registered (and fault prone) data is backed with automatically registered
data, the indicators for the management information are a better representation of the truth.
Third, it provides information that is now hard to come by, because the patient journey continues
after hospitalization. The patient data that is created after hospitalization could be added. This will
support the ongoing development of (quality) indicators. In addition, this could result in
identification of certain patient types, thereby providing the possibility to improve patient care.
In turn, these identified patient types provide opportunities for research into, and application of,
personalized healthcare and forecasting. The results can be validated by monitoring the key
performance indicators thereafter.
In summary, all these aspects will result into improved healthcare, benefiting all aforementioned
stakeholders.
Acknowledgements UMC Utrecht
Keywords: FAIR, management information, quality indicators
Abstractbook Health‐RI conference 2020
Abstract: 14 (Poster & Pitch)
Use of encrypted BSN in record linkage of epidemiological cohorts and
biobanks with disease to ensure valid linkages with optimal privacy
protection. A.W. (Sandra) van den Belt‐Dusebout (1), R. (Rosemarie ) Wijnands (1), O. (Otto) Visser (2), H.
(Hannelore) Hofhuis (3), J.L. (Hans) van Vlaanderen (4), J.A. (Jasper) Bovenberg (5), G. (Gerard) van
Grootheest (6), F.E. (Floor) van Leeuwen (1)
(1) Antoni van Leeuwenhoek ‐ The Netherlands Cancer Institute (NKI‐AVL), Amsterdam, (2)
Netherlands Comprehensive Cancer Organisation (in Dutch, IKNL), Utrecht, (3) The nationwide
network and registry of histo‐ and cytopathology in the Netherlands (in Dutch, PALGA), Houten, (4)
ZorgTTP, Houten, (5) jurist, (6) GGZ inGeest, Amsterdam
Background and purpose
Record linkage between cohorts/biobanks and (disease) registries is essential to efficiently and
validly answer important epidemiologic research questions. But, in the Netherlands it is prohibited by
law to use the Citizen Service Number (in Dutch: Burger Service Nummer, BSN) for research
purposes. Therefore, record linkage can only be performed using personal identifying data, e.g.
name, date of birth and postal code. However, large registries also increase the chance of
incorrect/uncertain links while restrictions in the Dutch Personal Data Protection Act (Dutch: Wet
bescherming persoonsgegevens) complicate checking links, yielding incorrect research outcomes.
Therefore, this project aimed to develop an improved standard record linkage procedure through use
of irreversibly encrypted BSNs, to ensure valid linkages with optimal privacy protection.
Methods
We used the nationwide OMEGA‐cohort of 42,000 women treated for subfertility between 1980 and
2001. Detailed information is available on fertility treatments, life‐style factors and cancer diagnoses
(through record linkages with the Netherlands Cancer Registry and PALGA). A large biobank consists
of toenail clippings providing DNA and tumor tissue blocks of hormone‐related cancers providing
more specific phenotypic information. ZorgTTP performed pseudonymization to enable anonymized
linkage on our behalf.
Results
After having overcome many procedural and ethical problems, several pseudonomized PALGA record
linkages and NCR record linkages based on encrypted personal identifiers have been performed.
Linkage results from NCR and PALGA were compared based on OMEGA identification numbers. The
best sensitivity and specificity combination for several record linkage scenarios based on personal
identifiers compared with the linkage based on encrypted BSN, was 0.91 and 0.98 and the worst
sensitivity and specificity combination was 0.91 and 0.77.
Conclusion
Using encrypted BSNs yields the best linkage results with optimal privacy protection.
Acknowledgements This study is granted by BBMRI‐NL 1.0 and BBMRI‐NL 2.0 Science Voucher. Tissue blocks have been
collected through BBMRI CP2011‐39. We thank VUmc and UMCU for providing encrypted BSNs.
Keywords: Record linkage; encrypted BSN; validation; privacy protection
Abstractbook Health‐RI conference 2020
Abstract: 15 (Poster)
The Ten Commandments of Translational Research Informatics Tim Hulsen (1)
(1) Philips Research
Translational research applies findings from basic science to enhance human health and well‐being.
In translational research projects, academia and industry work together to improve healthcare, often
through public‐private partnerships. This “translation” is often not easy, because it means that the
so‐called “valley of death” will need to be crossed: many interesting findings from fundamental
research do not result in new treatments, diagnostics and prevention. To cross the valley of death,
fundamental researchers need to collaborate with clinical researchers and with industry so that
promising results can be implemented in a product. The success of translational research projects
often does not depend only on the fundamental science and the applied science, but also on the
informatics needed to connect everything: the translational research informatics. This informatics,
which includes data management, data stewardship and data governance, enables researchers to
store and analyze their ‘big data’ in a meaningful way, and enable application in the clinic. The author
has worked on the information technology infrastructure for several translational research projects
in oncology for the past nine years, and presents his lessons learned in this poster in the form of ten
commandments. These commandments are not only useful for the data managers, but for all
involved in a translational research project. Some of the commandments deal with topics that are
currently in the spotlight, such as machine readability, the FAIR Guiding Principles and the GDPR
regulations. Others are mentioned less in the literature, but are just as crucial for the success of a
translational research project.
Acknowledgements The author would like to thank everyone involved in the CTMM‐TraIT, CTMM‐PCMM, Movember
GAP3, ERSPC, RE‐IMAGINE and LIMA projects.
Keywords: Translational research, medical informatics, data management, data curation, data
science
Abstractbook Health‐RI conference 2020
Abstract: 16 (Poster)
A tranSMART driven clinical diagnostic decision support system on
anemia J. Worst DBA (1) and Prof. Dr. H.J. van den Herik (1)
(1) Leiden University
Data integration
In Health‐RI, the common goal is to interconnect the biomedical resources, empowering researchers
to develop better personalized medicine and health solutions. A tranSMART driven CDDSS focused
on anemia will in the context of personalized medicine support a correct prognosis, which is
regarding the current available medical knowledge possible. Cooperation focused on the integration
of clinical data leads according to Engelen (2018) to medical knowledge, which at present (2019)
doubles every 3 to 4 years.
Clinical data patterns representing the health condition of elderly patients
Signs and symptoms of a mild or moderate anemia are often asymptomatic, e.g., breathlessness
and/or fatigue upon a strenuous exercise, which are common in severe anemia. Anemia (Bunn and
Aster, 2011; Boogaerts and Verhoef, 2017) is based on thinking about production versus destruction
of red blood cells. It explains the level of circulating red blood cells. The erythropoiesis is an example
of a complex system. Ineffective erythropoiesis is critical to the pathophysiological explanation of
destructive anemia such as iron deficiency, myelodysplastic, and megaloblastic.
Our study detected data patterns of an anemia underlying disease as a clinical indicator. Anemia
reflects its influences on the immunity system of elderly (those > 65 years), which makes them
vulnerable for acute disease. The traditional notion has been that anemia in elderly individuals
always reflects a serious underlying condition. It has long been recognized that a proportion of
patients, usually elderly, have anemia that does not meet diagnostic criteria for a specific etiology
(unexplained anemia), which concerns a prevalence of 17 to 45 % among elderly. At present the
ending of the life‐span after 20 years or more of a 65 year old is a reality.
Acknowledgements We like to thank Lygature, Health‐Ri and the Hyve
Keywords: anemia, tranSMART, CDDSS, elderly patients, data patterns, life‐span
Abstractbook Health‐RI conference 2020
Abstract: 17 (Poster)
Data Stewardship Wizard Rob Hooft (1), Marek Suchánek (2), Vojtěch Knaisl (2), Jan Slifka (2), Robert Pergl (2)
(1) DTL, The Netherlands, (2) Czech Technical University in Prague, Czech Republic
We will present the latest state of our tool, the Data Stewardship Wizard.
The Data Stewardship Wizard is a tool for data management planning that is focused on getting the
most value out of data management planning for the project itself rather than on fulfilling
obligations. It is based on FAIR Data Stewardship, in which each data‐related decision in a project
acts to optimize the Findability, Accessibility, Interoperability and/or Reusability of the data. The
background to this philosophy is that the first reuser of the data is the researcher himself. The tool
encourages the consulting of expertise and experts, can help researchers avoid risks they did not
know they would encounter, and can help them discover helpful technologies they did not know
existed.
Data management planning has several sociological problems:
The activity is seen as an obligation, a burden, by researchers.
Some data management novices underestimate the risks of insufficient data management and
data management planning and think that their knowledge of computing in the home
environment extrapolates to research data management in the lab.
It is hard for experts in specific aspects of data management to be found by the researchers that
need them the most. Expertise lists do not work for users who are unaware that they will be
running into a specific data management problem during the project.
We try to solve those problems using our tool, the “Data Stewardship Wizard”. We use the term
“Data Stewardship” to indicate that the activity is not only taking place during the project, but
extends to the long term maintenance of the resulting research data. We use the term “Wizard” to
refer to the tool as an “expert system” providing context dependent guidance to its users.
Our wizard:
alleviates the negative view of data management planning by focusing primarily on the benefits
for the research project itself and the researcher, not on the obligations;
can help to show researchers all the different aspects of data management: IT, archival,
sustainability and the entire FAIR data spectrum. The guidance tells stories of experts who have
learned their lessons the hard way;
points to available experts and expertise exactly where the issue at hand is brought up in the
questionnaire.
The Data Stewardship Wizard (https://ds‐wizard.org/)
presents questions in a hierarchical fashion, so that only relevant data management subjects are
presented to the user;
can function as a checklist for data stewards operating in a project, just like pilots use a checklist
to fly a plane: it ascertains that the experts do not forget any aspects of the planning;
Abstractbook Health‐RI conference 2020
consists mostly of closed questions, encouraging thinking through all aspects and avoiding the
problem that the researcher does not know where to start writing, thereby preventing the urge to
copy an existing data management plan from another project.
Technically, the wizard consists of
an open source web tool (with containerized installation) to present hierarchical data
management questionnaires, storing intermediate results in a database;
a knowledge model that contains a few hundred questions and is easy to extend;
a system to maintain knowledge models and to adapt them to your own institute or
infrastructure.
a templating engine that can be used to transform the data stewardship plan into a standard
data management plan following funder guidelines.
Acknowledgements This work has been paid partly by ELIXIR, with in‐kind contributions by DTL and Czech Technical
University in Prague
Keywords: DMP, Data Management
Abstractbook Health‐RI conference 2020
Abstract: 18 (Poster)
Data Desks at the University Medical Centers: Facilitating Access to
Expertise on Data Handling Rob Hooft (1), Anne‐Lotte Masson (2), Margo van Reen (3), Mirjam Brullemans (3), Erik van Iperen
(4), Rudy Scholte (4), Harry Pijl (5), Judith Manniën (6), Petra van Overveld (6), Pascal Suppers (7),
Ronald van Schijndel (8) and Salome Scholtens (9)
(1) DTL, (2) Erasmus MC, (3) Radboudumc, (4) Amsterdam UMC locatie Meibergdreef, (5) UMC
Utrecht, (6) LUMC, (7) MUMC+, (8) Amsterdam UMC locatie de Boelelaan, (9) UMCG
Each of the University Medical Centers (UMCs) in The Netherlands has its own data expertise center.
These provide help with many different aspects of data management to all researchers. Over the last
years we have brought these groups together in data4lifesciences work package "Access to
Expertise” and we will keep coming together inside Health‐RI in the future.
The group collaborates in order to:
* Exchange experience on how to run this kind of expertise desks (organisation and business)
* Exchange experience of the kind of questions received and how they are answered
* Discuss the position of data stewards in the UMCs
* Exchange practices for training of researchers as well as support staff
* Discuss related initiatives in The Netherlands, e.g. LCRDM
Through the exchange of information we want to achieve a landscape in which researchers from any
of the organisations can get access to expertise in the entire network.
Acknowledgements NFU
Keywords: data stewardship, expertise, support, NFU, collaboration
Abstractbook Health‐RI conference 2020
Abstract: 19 (Poster)
Changes in the intracranial volume from early adulthood to the sixth
decade of life Yaron Caspi (1), Rachel M. Brouwer (1), Hugo G. Schnack (1), Marieke E. van de Nieuwenhuijzen
(1), Wiepke Cahn (1), Renè S. Kahn (1,2), Wiro J. Niessen (3), Aad van der Lugt (3), Hilleke Hulshoff
Pol (1)
(1) UMC Utrecht Brain Center, Department of Psychiatry, University Medical Center Utrecht, The
Netherlands, (2) Department of Psychiatry, Icahn School of Medicine at Mount Sinai, New York, NY,
USA, (3) Department of Radiology and Nuclear Medicine, Erasmus MC: University Medical Center
Rotterdam, The Netherlands
Aging is manifested in structural changes of the brain. To understand normal and abnormal brain
aging, it is important to study various brain biomarkers among different age groups.
The total intracranial volume (ICV) is an important biomarker. Previous work about ICV aging
gravitated around two views. The first view claims that a substantial ICV reduction of about 0.2
%/year occurs during adulthood. The second view suggests that the ICV stays constant during
adulthood. In light of these conflicting positions, it is essential to clarify if, and to what extent, ICV
changes with age.
We measured IC in MRI (T1w) brain scans using a longitudinal design. Subjects were scanned at three
different time points with an average time interval of 3 years between scans (number of individuals
was 563, 363 and 323; their mean ages were 27.13±7.23, 30.07±7.15 and 33.67±7.57 years). By
applying a semi‐automatic in‐house‐build algorithm for IC volume extraction, we measure individual
trajectories of ICV changes between 20 and 62 years. This procedure allows us to detect within‐
individual changes in IC with increasing age.
Using three different analysis methods, we detected subtle but statistically significant longitudinal
trajectories of the ICV from adulthood to the middle of the sixth decade of life. Though the extent of
change differ between the analysis methods, they all show the same trend and order‐of magnitude
changes. E.g., one of the methods show that at age 20 there was an increase of +0.03 %/year, and at
age 55 there was a decline of ‐0.09 %/year. Thus, ICV changes from positive growth to negative
decline and is accelerates with age.
When using a cross‐sectional approach, we find a constant ICV decline rate of about 0.2 %/year.
Thus, the cross‐sectional approach estimated a stronger decline than the cross‐sectional one that we
interpret as a generational effect.
Acknowledgements The authors would like to thank Hakim Achterberg, Marcel Koek, Adriaan Versteeg, Thomas Phil,
Thomas Kroes, Baldur van Lew, Marcel Zwiers and Seyed Mostafa Kia for a collaboration within the
BBMRI‐NL work‐package 3.
This work was supported by the Netherlands Organization for Scientific Research (NWO
184.033.111), Biobanking and BioMolecular resources Research Infrastructure The Netherlands
(BBMRI‐NL2.0), and by the ENIGMA World Aging Center grant (NIH 1R56AG058854‐01, subaward
112068003).
Abstractbook Health‐RI conference 2020
The infrastructure for the GROUP study is funded through the Geestkracht programme of the Dutch
Health Research Council (Zon‐Mw, grant number 10‐000‐1001), and matching funds from
participating pharmaceutical companies (Lundbeck, AstraZeneca, Eli Lilly, Janssen Cilag) and
universities and mental health care organizations (Amsterdam: Academic Psychiatric Centre of the
Academic Medical Center and the mental health institutions: GGZ Ingeest, Arkin, Dijk en Duin, GGZ
Rivierduinen, Erasmus Medical Centre, GGZ Noord Holland Noord. Groningen: University Medical
Center Groningen and the mental health institutions: Lentis, GGZ Friesland, GGZ Drenthe, Dimence,
Mediant, GGNet Warnsveld, Yulius Dordrecht and Parnassia psycho‐medical center The Hague.
Maastricht: Maastricht University Medical Centre and the mental health institutions: GGzE, GGZ
Breburg, GGZ Oost‐Brabant, Vincent van Gogh voor Geestelijke Gezondheid, Mondriaan, Virenze
riagg, Zuyderland GGZ, MET ggz, Universitair Centrum Sint‐Jozef Kortenberg, CAPRI University of
Antwerp, PC Ziekeren Sint‐Truiden, PZ Sancta Maria Sint‐Truiden, GGZ Overpelt, OPZ Rekem. Utrecht:
University Medical Center Utrecht and the mental health institutions Altrecht, GGZ Centraal and
Delta.
Keywords: Intracranial Volume, Aging, Brain age, MRI, Longitudinal design, Cross‐sectional design.
Abstractbook Health‐RI conference 2020
Abstract: 20 (Poster)
The PALGA Portal ‐ Streamlining and professionalizing the request,
delivery and use of pathology data and materials. Annette Bruggink (1), Rinus Voorham (1), Stefan Willems (2) , Iris Nagtegaal (3), Folkert van
Kemenade (4)
(1) PALGA, (2) UMCU, (3) RadboudUMC, (4) ErasmusMC
PALGA, the Dutch pathology registry, delivers data from national databases for purposes such as
scientific research, medical quality control, and for the evaluation and monitoring of screening
programs. Furthermore PALGA offers the option of linking cohort data via a Trusted Third Party (TTP).
PALGA contains over 85 million pathology records and the accompanying materials, FFPE blocks and
tissue slides, are stored in 45 pathology labs. The PALGA Portal allows fast, easy and safe access to
these resources and with that, stimulates secondary use of pathology data and tissues for research.
The PALGA Portal was built in collaboration with BBMRI. The PALGA portal is a web‐based portal that
allows researchers to request pathology data or material from almost all diagnostic pathology labs in
the Netherlands. Laboratory Requests are forwarded to the designated labs and track‐and‐traced.
HUB‐employees, stationed in every academic hospital and serving the non‐academic labs, aid in
picking, registering and sending the requested materials.
Before the start of the PALGA portal almost all pathology labs were visited to introduce the PALGA
portal. In 2018 35 of 45 pathology labs were visited to evaluate the use of the PALGA portal. We have
spoken with more than 100 pathologists and laboratory staff to give an update about the PALGA
portal, to discuss the changes under the GDPR in pathology research, and to retrieve any
improvements.
In 2019 (Q1‐Q3) 144 requests for ‘PA material’ were send to the laboratories. 19.558 PA numbers
were requested from which 14.745 consists of FFPE material. The other 4.813 where Pathology
reports or clinical data.
The PALGA Portal has streamlined and professionalized the request, delivery and use of pathology
data and material for research. It has shown to increase efficiency and transparency for both the
requesting researchers and providing pathology labs.
Acknowledgements BBMRI
Keywords: PALGA portal, Data, FFPE blocks
Abstractbook Health‐RI conference 2020
Abstract: 21 (Poster)
Generating CT‐scans with 3D Generative Adversarial Networks Using
Supercomputers (1) David Ruhe, (2) Valeriu Codreanu, (3) Caspar van Leeuwen, (4) Damian Podareanu, (5) Vikram
Saletore, (6) Jonas Teuwen
(1) SURFsara, (2) Intel, (3) NKI
It is already a well‐known fact in the computer vision community that current deep learning methods
achieve very accurate results compared to traditional methods. These approaches are very data
hungry, and also require that training data overlaps as much as possible with the true data
distribution, such that the algorithmic bias is minimized when deploying these systems in the real‐
world. Although deep learning applied to medical imaging has shown to achieve good results within
the same medical centre that provided the training data, generalization to other centres is often poor
because of a lack of large, multi‐centre (public) datasets. One of the main reasons that such data sets
are scarce is that privacy concerns make sharing very difficult. To overcome this challenge, this study
aims to generate synthetic 3D CTs (computed tomography), that would allow hospitals around the
world to share medical images that follow the same data distribution as their real.
We extend recent works that develop a technique that progressively grows GANs (Generative
Adversarial Networks) during training to voxel space, and validate our techniques using the open
LIDC‐IDRI thoracic CT dataset. Generating CT samples is very challenging in terms of computational
and memory requirements, as it requires both using 3D convolutional layers, as well as the ability to
generate large‐dimensionality 512x512x128 scans. In order to iterate faster over this high
computational complexity model, we have used distributed training, our current models being
trained on up to 256 dual‐socket Intel Cascade Lake nodes (~12000 cores).
Acknowledgements We would like to acknowledge the Endeavor support team
and especially Mr. Mallick Arigapudi.
Keywords: medical imaging; deep learning; generative adversarial networks
Abstractbook Health‐RI conference 2020
Abstract: 22 (Poster)
Medical images and AI: the need of a big data revolution Alberto Traverso (1), Ivan Zhovannik (1, 2), Ibrahim Hadzic (1), Suraj Pai (1), Dominik Jeurissen (1),
Andre Dekker (1)
(1) Department of Radiotherapy, Maastro Clinic, Doctor Tanslaan 12, 6229ET Maastricht (NL), (2)
Department of Radiotherapy, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525GA
Nijmegen (NL)
Medical images represent the larger percentage of data produced in the clinic with a key role in
radiation oncology. Cancer diagnosis, treatment and monitoring are based on patients’ scans.
However, visual inspection of medical images does not catch the unique information that scans
embed. This is because medical images should be considered more as just pictures, they are big data.
The advent of AI (Artificial Intelligence)‐driven computational pipelines opened the possibility to
automatically inspect images via “artificial eyes” and extract valuable information, more than as
humans we can process, this is called quantitative imaging. This information has the advantage to
non‐invasive (compared for example to biopsies) exploit the unique biology of a patient tumor and it
can support data driven decisions about our patients: better decisions, better cancer care. However,
the medical imaging community was not ready for this revolution. After an initial phase of
excitement with promising results for several diseases, investigation of the weakness and drawbacks
of this new methodology followed especially when it exists a gap between academic results and
translation in the clinic. Lack of transparency, low quality of input data, absence of accepted
methodology, complexity of the problem to be solved, generalizability of image‐derived models
represents the major drawbacks. Practically, all these drawbacks were generated by not considering
medical images and derived information as big data, with all the issues that come with that, but also
with all the methodology already available to tackle this hyperdimensional problem. Only by taking
steps back and recognizing the importance of considering quantitative imaging as a big data problem
we could solve the above‐mentioned issues. We will present our ongoing efforts to revitalize
quantitative imaging, which include: improvement of images’ quality, better and reproducible AI
tighten with FAIR principles and common computational infrastructure that are strong than data‐
sharing barriers.
Acknowledgements The authors would like to acknowledge their colleagues at Moffitt Cancer Centre, Princess Margaret
Cancer Centre and the DIAG (Diagnostic Image Analysis Group) in Nijmegen.
Keywords: medical imaging, AI, big data, transparent science, radiation oncology
Abstractbook Health‐RI conference 2020
Abstract: 23 (Poster)
Laboratory variation of molecular testing in a Dutch cohort of
metastatic non‐small cell lung cancer patients from 2017 (1) Betzabel Cajiao Garcia, (2) Chantal CHJ Kuijpers, (3) Michel M van den Heuvel, (4) Anne SR van
Lindert, (5) Ronald AM Damhuis, (6) Stefan M Willems
(1) University Medical Centre Utrecht, (2) University Medical Centre Utrecht, Foundation PALGA, (3)
Radboudumc, (4) University Medical Centre Utrecht, (5) Netherlands Comprehensive Cancer
Organisation, (6) University Medical Centre Utrecht
Background & objective: Adequate and timely testing for genetic alterations in non‐small cell lung
cancer (NSCLC) is necessary to consider targeted therapy when a certain genetic alteration is present.
Previously, we demonstrated that in the Netherlands molecular testing was suboptimal in 2015, as
25% (EGFR/KRAS and ALK) to 50% (ROS1) of patients were not tested according to guidelines, and
notable laboratory variation was present. Currently, by analyzing a cohort of metastatic NSCLC from
2017, we aim to assess whether the performance of molecular testing improved.
Methods: All stage IV non‐squamous NSCLC with incidence year 2017 from the Netherlands Cancer
Registry were matched to the Dutch pathology registry (PALGA). Using information extracted from
pathology excerpts, proportions of tumors tested for EGFR/KRAS mutation, and ALK, ROS‐1, and RET
rearrangement <3 months after diagnosis were determined and variation between 42 laboratories
was assessed.
Results: Of 3746 identified patients, we have currently analyzed 1565 (42,0%). Fifty‐five patients
were non‐eligible after matching, leaving 1510 (40,3%) patients. EGFR/KRAS testing was performed
in 1245 patients (82.5%) (laboratory variation 63.6‐100%). Of the EGFR/KRAS wildtype tumors
(n=608), 422 (69.4%) tumors were tested for ALK (33.3‐100%), 305 (50.2%) for ROS‐1 (0‐100%), and
214 (35.2%) for RET (0‐100%). Insufficient tumor tissue and inappropriate specimen were the most
stated reasons for not testing.
Conclusions: These preliminary data show significantly higher EGFR/KRAS, ALK and ROS‐1 testing
proportions compared to 2015. Further improvement remains possible, in some laboratories more
than in others, and especially for ROS‐1 and RET testing, to identify candidates for targeted therapy.
Combining FAIR data from two national databases facilitates data‐driven feedback of clinicians to
enhance personalized treatment of lung cancer patients.
Acknowledgements UMC Utrecht, Foundation PALGA, Pfizer, Roche, AstraZeneca
Keywords: Non‐Small‐Cell Lung Carcinoma, molecular testing, personalized medicine, data
agregation, PALGA, FAIR data
Abstractbook Health‐RI conference 2020
Abstract: 24 (Poster)
Herziening Gedragscode gezondheidsonderzoek Dr. Martin Boeckhout (1), mr. Evert‐Ben van Veen (1)
(1) MLC Foundation
De Gedragscode gezondheidsonderzoek uit 2004 heeft geruime tijd duidelijkheid geboden over de
voorwaarden om patiëntgegevens in gezondheidsonderzoek te mogen gebruiken. Inmiddels is er de
AVG met helaas ook soms de ‘AVG kramp’ en is ook het gezondheidsonderzoek aanzienlijk van
karakter veranderd. Met steun van VWS en ZonMw is COREON de herziening van de Gedragscode
gestart. De praktische uitvoering is belegd bij de MLC Foundation. In tegenstelling tot de vorige zal de
herziene Gedragscode handvaten bieden voor alle gegevensverwerking bij gezondheidsonderzoek,
bijvoorbeeld ook rond gegevens die in het kader van een WMO studie, van vrijwilligers aan nWMO
onderzoek of via biobanken worden verkregen. De poster beschrijft:
• de aanleiding(en) voor de herziening;
• het proces van herziening en op welke wijze de diverse stakeholders daarbij worden
betrokken;
• voorbeelden van de onderwerpen die in de Gedragscode aan de orde zullen komen;
• de opbouw van de Gedragscode;
• de tijdlijn.
Acknowledgements ZonMw
Keywords: AVG, GDPR, WMO, nWMO,
Abstractbook Health‐RI conference 2020
Abstract: 25 (Poster)
Supporting your Research; Tools for Data Management and Processing Project and Steering Committee Research ICT (1)
(1) Amsterdam UMC, Amsterdam, The Netherlands
In 2018, an ambitious four‐years plan was launched, aiming to boost and harmonize data
management and IT support for researchers within Amsterdam UMC. Here is an overview of services,
including dedicated workspaces, that recently already have become available for researchers:
Data Management Support –A newly established helpdesk can help to review data management
plans and provides support on data collections and applications.
Research Data Platform – Clinical data from Amsterdam UMC patients gathered during care though
Epic and other sources are currently collected in a research data platform and can be extracted in a
GDPR‐compliant format – anonymous, encoded or identifiable, depending on the legal requirements.
CTcue Patientfinder – CTcue uses Epic to search for patients fitting study criteria based on up‐to‐date
structured (i.e., diagnose, medication, lab) as well as unstructured (i.e., notes, reports, letters) data.
Castor Campus License – Castor enables researchers to easily capture and integrate data in a GDPR‐
compliant manner. A new campus license offers the Castor eCRF free of charge for non‐commercial
research activities at Amsterdam UMC.
Azure‐based DRE – The anDREa consortium led by Radboudumc is developing a user‐friendly digital
research environment where researchers can work with all their data, analytics, and tools in a secure
and self‐serviced manner. Researchers from Amsterdam UMC have access to this environment.
Research Cloud – The Amsterdam UMC Research Cloud is a platform for flexible and advanced
computing, hosted within the SURFcumulus research environment. The cloud is designed for and
available for IT‐experienced researchers
Research Zone – The research zone is a network specifically set up for research and separated from
other networks, like the care domain network. The research zone facilitates connections to
(inter)national networks and shared resources, including data, storage, software, and high‐
performance computing.
Acknowledgements None
Keywords: Data management Support, Research Data Platform, CTcue, Castor, anDREa, Research
Cloud
Abstractbook Health‐RI conference 2020
Abstract: 26 (Poster)
FAIR Genomes: Standardizing a meta‐data schema for FAIRifying
personal genome data workflows Gurnoor Singh (1) , K. Joeri van der Velde (2), Jeroen Beliën (4), Jasmin Böhmer (3), Daphne
Stemkens (5), Lisenka Vissers (1), Jeroen van Reeuwijk (1), Saskia Hiltemann (7), Lennart Johansson
(2), Nienke van der Stoep (6), Daoud Sie (4), Janneke Weiss (4), Geert Frederix (3), Marco Roos (6),
Erik van Iperen (8), Terry Vrijenhoek (3), Folkert W. Asselbergs (3), Joris van Montfrans (3), Rolf
Sijmons (2), Hanneke van Deutekom (3), Pieter Neerincx (2), Joep de Ligt (3), Fernanda de Andrade
(2), Anna Niehues (1), Hindrik H.D. Kerstens (10), Mark Thompson (6), Rajaram Kaliyaperuma (6),
Annika Jacobsen (6), Katy Wolstencroft (6, 14), Ies Nijman (3), Marcel Nelen (1), Ariaan Siezen (1),
Koen ten Hove (1), Nine Knoers (2), Christian Gilissen (1), Hans Scheffer (1), Stefan Willems (3),
Wendy van Zelst‐Stams (1), Helger IJntema (1), Kim Elsink (3), Bart de Koning (9), Bauke Ylstra (4),
Erik Sistermans (4), Patrick Kemmeren (10), Henne Holstege (4), Christine Staiger (11), Bastiaan
Tops (10), Susanne Rebers (12), David van Zessen (7), Valesca Retèl (12), Edwin Cuppen (13), Peter
van Tintelen (3), David van Enckevort (2), Lieneke Steeghs (1), Salome Scholtens (2), Jeroen Laros
(6), Leon Mei (6), Cor Oosterwijk (5), Andrew Stubbs (7), Peter A.C. ‘t Hoen (1), Mariëlle van Gijn
(2), Morris Swertz (2)
(1) Radboud University Medical Center, Nijmegen, The Netherlands, (2 ) University Medical Center
Groningen, The Netherlands, (3) University Medical Center Utrecht, The Netherlands, (4) Amsterdam
University Medical Centers, location VUmc, The Netherlands, (5) = VSOP ‐ Dutch Patient Alliance for
Rare and Genetic Diseases, (6) Leiden University Medical Center, The Netherlands, (7) Erasmus
Medical Center, Rotterdam, The Netherlands, (8) Durrer Center for Cardiovascular Research, Utrecht,
The Netherlands, (9) Maastricht University Medical Center, The Netherlands, (10) Princess Máxima
Center for Pediatric Oncology, Utrecht, The Netherlands, (11) Dutch Techcentre for Life Sciences,
Utrecht, The Netherlands, (12) Netherlands Cancer Institute, Amsterdam, The Netherlands, (13)
Hartwig Medical Foundation, Amsterdam, The Netherlands, (14) Leiden Institute for Advanced
Computer Science, Leiden University, Leiden, The Netherlands
The increase in personal genome data generated in diagnostics and research holds great promise for
advancing personalized prevention and medicine. However, valuable genomic and associated clinical
data is fragmented across many healthcare providers and research organizations, making it difficult
to reuse due to lack of findability, accessibility and interoperability. This prohibits us from exploiting
the potential information contained in these genomes for health benefit. FAIR Genomes aims to
provide guidelines that should increase reuse of genomic data while considering the needs of all
stakeholders and addressing ELSI issues.
We present a standardized meta‐data schema to harmonize genomic data workflows and their
reporting practices. This schema is broadly segmented into five categories: general information;
informed consent; personal and clinical information; material information and technical information.
In face‐to‐face and videoconference meetings, we work towards defining the schema, which is a list
of common and optional data elements with relationships and values mapped to existing ontologies
such as SNOMED, DUO, HPO, UMLS and EDAM. This project aims to make all data and meta‐data
elements findable and interoperable to increase FAIRness and standardization in capturing genomic
data. This meta‐data schema provides a strong basis for digital twin data in Dutch hospitals,
development of personal genetic lockers, and active Dutch participation in the European '1+ Million
Genomes' Initiative.
Abstractbook Health‐RI conference 2020
The scope of this schema goes beyond to next‐generation DNA sequencing data. We expect to
expand into various *omics varieties, as well as capturing analysis pipelines in FAIR terms. Hence, the
FAIR Genomes meta‐data framework could be used to develop other research‐based infrastructures
such as X‐omics, BBMRI, ELIXIR, Solve‐RD and European Joint Programme on Rare Diseases.
The FAIR Genomes meetings are open to receive input from anyone to achieve the highest quality
and usability of the resulting meta‐data framework. Join us at: https://github.com/fairgenomes.
Keywords: FAIR, datasharing, genomics, healthcare, meta‐data, ontologies, framework
Abstractbook Health‐RI conference 2020
Abstract: 27 (Poster)
BBMRI‐Omics: Valuable resource of multi‐omics data and analysis
tools Marian Beekman (1), Jurriaan Barkey Wolf (1), Davey Cats (1, 2), Joyce van Meurs (3), Lude Franke
(4), Bastiaan T. Heijmans (1), Morris Swertz (4), Leon Mei (1, 2), Cornelia van Duijn (5), Dorret I.
Boomsma (6), P. Eline Slagboom (1), GONL (7), BIOS Consortium (8), BBMRI Metabolomics
Consortium (9)
(1) Molecular Epidemiology, LUMC, Leiden, (2) Sequencing Analysis Support Core, Leiden University
Medical Center, Leiden, The Netherlands, (3) Internal Medicine, Erasmus University Medical Center,
Rotterdam, (4) Genetics, University Medical Center Groningen, University of Groningen, Groningen,
The Netherlands, (5) Epidemiology, Erasmus University Medical Center Rotterdam, Rotterdam, The
Netherlands, (6) Biological Psychology, Vrije Universiteit, Amsterdam, The Netherlands, (7) BBMRI‐NL
consortium Genome of the Netherlands, (8) BBMRI‐NL consortium Biobank‐based Integrative Omics
Studies, (9) BBMRI‐NL consortium Metabolomics
BBMRI‐Omics is the joint collection of omics data that has collaboratively been generated on
thousands of participants of 29 Dutch biobanks and that is made available for BBMRI researchers
focusing on integrative omics studies. BBMRI‐Omics is publicly available and has proven to facilitate
researchers in their discovery of novel biological mechanisms and biomarkers for health and disease.
BBMRI‐Omics consists 4,000 individuals with integrative genomics data (genome, epigenome,
transcriptome and metabolome) with an extension of the metabolome in 30,000 extra individuals
and whole genome sequences in a selective group of 700 individuals. BBMRI‐Omics also provides
tools (on gitlab and Bioconductor) and computational space both facilitating omics data analysis. The
summary statistics of cross‐omics association analyses, like expression QTLs, methylation QTLS,
metabolite QTLs are browsable in the BBMRI‐Omics Atlas (bbmri.researchlumc.nl/atlas), where for
example genes or genetic locations of interest can be browsed for association with DNA methylation,
metabolite levels and/or gene expression in blood. Soon it will be possible to link available omics
data to brain available imaging. The unique scale of the BBMRI‐Omics infrastructure, in the sense of
the number of individuals as well as the amount of data per individual, enables researchers to
investigate a broad spectrum of research questions.
Acknowledgements This work was financially supported by BBMRI‐NL, a Research Infrastructure financed by the Dutch
government (NWO, numbers 184.021.007 and 184.033.111).
Keywords: Whole genome sequences, transcriptome, methylome, metabolome, multi‐omics data,
analysis tools
Abstractbook Health‐RI conference 2020
Abstract: 28 (Poster)
Uitdagingen bij het bouwen van een Fair Data station Jack Broeren (1)
(1) stakeholders
We willen informatie delen over waar je tegen aan kan lopen als je een FAIR data station inricht voor
productiedoeleinden. Het is relatief simpel om in een POC of Pilot een klein demo‐project op te
zetten. Als je dit dan vervolgens wil uitbreiden naar full‐scale dan loop je tegen allerlei zaken aan als:
performance, inrichten laad‐processen, update‐mechanismes etc.
Acknowledgements stakeholders
Keywords: Big data; scaling;performance; architectuur;
Abstractbook Health‐RI conference 2020
Abstract: 29 (Poster)
EATRIS‐Plus ‐ a multi‐omic toolbox to support cross omic analysis and
data integration in clinical samples Ms Anne‐Charlotte Fauvel (1) Dr Florence Bietrix (1), Dr Andreas Scherer (2), Prof. Alain van Gool
(3), Prof. Peter‐Bram 't Hoen (3), Prof. Marian Hajduch (4), Dr. Antonio Andreu (1)
(1) EATRIS, (2) FIMM, (3) Radboudumc, (4) IMTM
Efficient advancement of Personalised Medicine depends on the availability of validated patient‐
targeted biomarkers. However, as our capacity to identify genetic variants associated with complex
diseases increases, these do not fully recapitulate the resulting disease phenotypes, and a more
precise understanding of the molecular profiles are needed. This realisation provides a rationale for
the development of multi‐omic approaches. In order to turn the multi‐omic promises into a reality,
systemic bottlenecks impacting the biomarker field needs to be overcome:
Poor levels of technological and analytical harmonisation;
Poor data stewardship and compliance to the FAIR (Findable, Accessible, Interoperable, and
Reusable) principles;
Lack of understanding of the relationship between genomic biomarkers and downstream
molecular markers (transcriptomic, proteomic, metabolomic, among others);
Lack of reliable control reference values for these biomarkers in healthy populations; and
Poor understanding of the clinical needs resulting in limited clinical adoption.
Tackling those issues in a systematic way is one of the objectives of EATRIS‐Plus, a H2020‐funded
project to kick start early 2020. With 19 partners across 13 countries, the consortium ambitions to
deliver a multi‐omic toolbox to support cross omic analysis and data integration in clinical samples.
This toolbox will contain:
Consensus‐based SOPs for omic technologies;
Guidelines for omic analytical processes;
Validated reference materials for analytical processes;
Quality parameters for benchmarking quality assessment activities;
Data analytical and FAIRification tools;
Criteria for establishing reference values in population cohorts;
Troubleshooting guidelines;
Access to a repository of multi‐omic reference values
The omic tools will be developed and tested with a real‐setting demonstrator, an already established
cohort of 1,000 healthy individuals in Czechia upon whom genomic sequencing has been already
performed. Information available on this healthy individual cohort will be augmented during the
project with transcriptomic, proteomic and metabolomic data.
By providing such toolbox to the research community, EATRIS‐Plus will be the engine to enable high‐
quality research in the context of patient stratification and accelerate the implementation of
Personalised Medicine solutions.
EATRIS is the European Infrastructure for Translational Medicine providing services for accelerating
biomedical innovation.
Abstractbook Health‐RI conference 2020
Acknowledgements This project has received funding from the European Union's Horizon 2020 research and innovation
programme under grant agreement No 871096
Keywords: personalised medicine, multi‐omics, FAIR data, patient stratification, translational
research
Abstractbook Health‐RI conference 2020
Abstract: 30 (Poster)
Fully automatic construction of optimal radiomics workflows Martijn P. A. Starmans (1, 2), Sebastian R. van der Voort (1, 2), Hakim Achterberg (1, 2), Marcel
Koek (1, 2), Razvan L. Micle (1), Milea J.M. Timbergen (3, 4), Melissa Vos (3, 4), Fatih Incekara (1, 5),
Maarten M.J. Wijnenga (6), Guillaume A. Padmos (1),
(1) Department of Radiology and Nuclear Medicine, Erasmus MC, Rotterdam, the Netherlands, (2) Department
of Medical Informatics, Erasmus MC, Rotterdam, the Netherlands, (3) Department of Surgical Oncology,
Erasmus MC, Rotterdam, the Netherlands, (4) Department of Medical Oncology, Erasmus MC, Rotterdam, the
Netherlands, (5) Department of Neurosurgery, Erasmus MC, Rotterdam, The Netherlands, (6) Brain Tumor
Center, Erasmus MC, Rotterdam, the Netherlands, (7) Department of Pathology, Erasmus MC, Rotterdam, the
Netherlands, (8) Faculty of Applied Sciences, Delft University of Technology, the Netherlands
Purpose: Radiomics uses combinations of imaging features to predict clinically relevant data. Many
radiomics methods have been described in the literature; however, there is no single method that
works for many applications. We present a Workflow for Optimal Radiomics Classification (WORC),
an open‐source solution to fully automatically construct an optimal workflow per application.
Methods and Materials: WORC states radiomics as a modular workflow, including multiple
algorithms and their parameters for each component. During training, WORC automatically adapts
itself by testing thousands of pseudo‐randomly defined radiomics workflows. The best workflows are
combined into one optimal signature. To evaluate WORC, three experiments on different clinical
applications were performed: (1) to classify 119 patients
with primary liver tumors in benign or malignant on T2‐weighted Magnetic Resonance Images scans
(MR); (2) to predict the 1p/19q co‐deletion in 287 patients with presumed low‐grade gliomas on T1‐
and T2‐weighted MR; and (3) to distinguish liposarcomas from lipomas in 88 patients on T1‐weighted
MR. Ground truth was obtained through pathology. Evaluation was implemented through a 100x
random‐split cross‐validation, with 80% of the
data used for training and 20% for testing. Performance is given in 95% Confidence Intervals (CIs).
WORC requires an efficient infrastructure to host these datasets, integrate different software
solutions and execute a large number of workflows. WORC therefore uses the fastr workflow engine
for managing the execution of automated analysis pipelines. Datasets are stored on XNAT and
experiments executed on the SURFSara Cartesius cluster, for which fastr contains plugins.
Results: The CIs of the area under the ROC curve were (0.86, 0.99) for liver tumors, (0.74, 0.85) for
brain tumors, and (0.74, 0.93) for lipomas/liposarcomas.
Conclusion: The results in three different applications demonstrate that WORC is a promising
approach for fully automatic construction of optimal radiomics workflows.
Acknowledgements Martijn Starmans acknowledges funding from the research program STRaTeGy (project number
14929‐14930), which is (partly) financed by the Netherlands Organisation for Scientific Research
(NWO). Sebastian R. van der Voort acknowledges funding from the Dutch Cancer Society (Koningin
Wilhelmina Fonds (KWF) project number EMCR 2015‐7859).
Keywords: workflow optimization, automatic algorithm selection, radiomics, machine learning,
oncology
Abstractbook Health‐RI conference 2020
Abstract: 31 (Poster)
Fastr workflow engine for reproducible and managed large‐scale
processing Hakim Achterberg (1), Marcel Koek (1), Adriaan Versteeg (1), Mahlet Birhanu (1), Martijn Starmans
(1), Thomas Kroes (2), Esther Bron (1), Wiro Niessen (1, 3)
(1) Biomedical Imaging Group Rotterdam, Department of Radiology & Nuclear Medicine, Erasmus
Medical Center, Rotterdam, the Netherlands, (2) Division of Image Processing (LKEB), Department of
Radiology, Leiden University Medical Center, Leiden, the Netherlands, (3) Department of Imaging
Physics, Faculty of Applied Sciences, Delft University of Technology, Delft, the Netherlands
Within the life‐science domain, much of the processing is not a single executable that is run, but a
combination of many executables that need to be run in a specific environment. Traditionally, this
was handled with shell scripts. However, with the increasingly complex analyses and size of the data,
this solution has reached its limits. There is a strong trend towards distributed execution of pipelines
in processing environments such as a cluster or cloud, which requires special orchestration.
Workflow engines like Fastr formalize how data flows between processing steps. This helps allows for
validation of the workflow before and during execution.
Fastr has been designed with consolidated workflows in mind. To this end there are a number of
important features: 1) management of tool versions, 2) data‐provenance, 3) workflow and
intermediate results validation to pinpoint errors on occurence, and 4) visualization of the execution
of a workflow using PIM. Fastr allows the tracking and the use of different versions of tools, this is to
ensure reproducibility, also in the future when tools are updated. The provenance model ensures
that there is a complete audit trail for all processed data. The complete definition of tools allows the
system to check if the workflow is valid before run, comparing data types of connected steps. During
execution the system automatically checks if valid results have been created for each step to detect
errors early on instead of propagating them. Finally, Fastr has a plugin to submit progress of a
workflow run to Pipeline Inspection and Monitoring (PIM) to give a visual representation of the run
via a web interface. Fastr has been used for processing >105 imaging sessions, leading to >107 of jobs
being executed on a cluster. In conclusion, Fastr is a robust workflow system enabling reproducible,
managed workflows.
Acknowledgements BBMRI‐NL 2.0
Keywords: pipeline, processing, HPC
Abstractbook Health‐RI conference 2020
Abstract: 32 (Poster)
Quantitative Imaging Biomarker Storage and Compute Infrastructure Marcel Koek (1), Hakim Achterberg (1), Adriaan Versteeg (1), Mahlet Birhanu (1), Henri Vrooman
(1), Thomas Phil 1), Thomas Kroes (2), Aad van der Lugt (1), Wiro Niessen (1,3)
(1) Department of Radiology & Nuclear Medicine, Erasmus MC, Rotterdam, the Netherlands, (2)
Division of Image Processing (LKEB), Department of Radiology, Leiden University Medical Center,
Leiden, the Netherlands, (3) Department of Imaging Physics, Faculty of Applied Sciences, Delft
University of Technology, Delft, the Netherlands
For extracting Quantitative Imaging Biomarkers (QIBs) from population or cohort based imaging
studies, a storage and compute infrastructure is essential. To make these QIBs meaningful, they can
be related to other data in (biobank) repositories. For this, we developed an infrastructure where
standardized image‐analysis pipelines can be managed and manual annotations and inspections can
be performed on large medical imaging datasets. The infrastructure is currently being used in
multiple population imaging studies. The QIBs and processed data can be linked to study and genetic
data, creating more comprehensive biobank repositories.
The infrastructure can be divided into three main components:
* Medical imaging data storage using XNAT
* Compute infrastructure
* Data and workflow management services
The compute infrastructure is developed to work in cloud environments and HPC clusters. The Fastr
workflow engine and PIM inspection and monitoring are the key components. These tools interface
with each other through REST APIs. Fastr is able to interact with the cluster and cloud environments
for executing jobs and to data storage directly through extensions.
The data and workflow management services are a collection of tools and services to manage the
data flow, including manual interaction with the data, in an automated fashion. The key components
are the Study‐Governor for automatically managing the data flow, Task‐Manager for task based
manual interaction with the data, and the ViewR to interact with the data on XNAT by researchers
based on the tasks served by the Task‐Mmanager. These components interface with each other
through REST APIs.
Our infrastructure can greatly benefit personalized medicine by making pipelines for imaging
biomarker extraction available to researchers and clinicians. Additionally, we create a reference
database for different imaging biomarkers, which can be used to compare an individual against the
general population. This will enable improved re‐use of imaging data for diagnostics and prognostics.
Acknowledgements BBMRI‐NL2.0, EPI2, EuroBioImaging
Keywords: Quantitative Imaging Biomarker, IT infrastructure, Population Imaging
Abstractbook Health‐RI conference 2020
Abstract: 33 (Poster)
Streamlining manual tasks in large medical imaging studies Adriaan Versteeg (1), Hakim Achterberg (1), Mahlet Birhanu (1), Henri Vrooman (1), Marcel Koek
(1), Aad van der Lugt (1), Wiro Niessen (1,2)
(1) Department of Radiology & Nuclear Medicine, Erasmus MC, Rotterdam, the Netherlands, (2)
Department of Imaging Physics, Faculty of Applied Sciences, Delft University of Technology, Delft, the
Netherlands
Medical imaging studies, even with all the automated methods, still require manual work; to assure
the quality of input data (QA), to control the output quality of the automated pipelines (QC), and to
create annotations that can be used to develop Machine/deep Learning algorithms.
We developed two applications to streamline the manual tasks leading to increased productivity,
reproducibility and quality. These applications (ViewR and Task Manager) work together, the Task
Manager keeps track of the work to be performed and the ViewR is used by users to perform the
work.
This ensures that the tasks are performed by a specific person or by a person with a specific skill set
and that this person always has the correct images and tools available for these tasks.
Each task contains the location of the data and a ViewR template. The ViewR uses this template to
setup the layout, editor capabilities and electronic Case Report Form to be filled. Furthermore, to
improve ease of use, tasks are loaded by the press of a single button and all images are preloaded so
loading a new task takes seconds instead of minutes.
The combination of the Task Manager and the ViewR can be used for:
▪ QA/QC in large cohort studies
▪ Incidental Findings
▪ Manual annotations for Machine learning
▪ Inter/Intra rater comparison
The Task Manager, ViewR combinations have been successfully used in the Rotterdam Scan Study to
perform both incidental finding and QC on approximately 2000 subjects. It has been used for the
CVON Heart Brain Connection project to mark/outline Infarcts on 500+ subjects. It has been used for
inspection tasks for more than 50.000 scans.
Acknowledgements BBMRI‐NL2.0, EPI2, EuroBioImaging
Keywords: Population Imaging, Infrastructure, Machine learning
Abstractbook Health‐RI conference 2020
Abstract: 34 (Poster)
Linkage of Lifelines and PALGA data: Enhancing multidisciplinary
research Rinus Voorham PhD (1), Annette Gijsbers PhD (1), Aafje Dotinga PhD (2), Trynke de Jong PhD (2)
(1) PALGA, (2) Lifelines cohort study
Multiple longitudinal data‐collections exist in the Netherlands with different levels of maturity, either
developed specifically for scientific research purposes (such as Lifelines) or for patient care purposes
(such as PALGA). Data linkage between these collections on the individual level is a powerful tool to
combine general health and lifestyle information with specialized clinical results, creating new
possibilities for multidisciplinary research.
Lifelines is a large, population‐based cohort study and biobank including 167,000 participants in the
Northern part of the Netherlands, among which three‐generation families. Lifelines allows linkage
between its own data and datasets from other data registries, including PALGA, in order to facilitate
scientific research in the field of healthy ageing.
PALGA, the nationwide network and registry of histo‐and cytopathology in the Netherlands, contains
over 85 million pathology records, maligned and benign, from all Dutch pathology laboratories.
Furthermore PALGA is the linkage between the investigator and the laboratories in case of requests
for pathology materials.
Data linkage with respect to privacy
Lifelines, PALGA, and ZorgTTP composed pseudonomization and encryption procedures making
linkage through Lifelines‐ and PALGA personal identifiers possible compliant with the Dutch privacy
protection laws. The linked (FAIR) dataset consisting of PALGA and Lifelines personal data matched
at the individual level by ZorgTTP, via PALGA pseudonyms and devoid of any Personal Identifiers, is
available on request through the secure remote access environment of Lifelines.
Useful for multidisciplinary research
The linkage of the PALGA pathology dataset with the Lifelines data on health, lifestyle and
demographics enables 1) linkage on request, 2) continuous update of data and 3) required privacy
protection of participants and patients. Ultimately, the linkage of Lifelines and PALGA databases will
strongly stimulate (prospective) multidisciplinary research aiming at personalized medicine and
health, leading to improvements in health care and disease prevention.
Acknowledgements ZorgTTP
Keywords: Data linkage Lifelines PALGA multidisciplinary research
Abstractbook Health‐RI conference 2020
Abstract: 35 (Poster)
Towards FAIR Data Steward as profession for the Lifesciences Salome Scholtens (1), Mijke Jetten (2), Jasmin Böhmer (3), Christine Staiger (4), Inge Slouwerhof
(2), Marije van der Geest (1), Margreet Bloemers (5), Ingeborg Verheul (6), Celia W.G van Gelder (4)
(1), Genomic Coordination Centre, UMCG, Groningen, (2) Radboud University Nijmegen, (3) UMC
Utrecht, (4) Dutch Techcentre for Life Sciences (DTL), Utrecht, (5) ZonMw, (6) LCRDM
Data stewardship expertise is essential in research. However, the lack of consensus on the function
profile of data stewards hampers adequate data steward capacity building in organisations. In our
ZonMw funded project entitled “Towards FAIR Data Steward as profession for the
Lifesciences”(Aug18‐Jul19), we delivered community‐endorsed job descriptions (including
responsibilities and tasks) and an agreement on the required knowledge, skills and abilities (KSAs) for
data stewards. To be able to build tailored data stewardship training, also detailed learning
objectives were formulated.
Our analysis of the data stewardship landscape shows three different, partly overlapping, data
stewardship roles which all have their own focus: policy, research and infrastructure
(https://doi.org/10.5281/zenodo.3243909). Furthermore we have identified 8 competence areas for
the data steward: Policy/strategy, Compliance, Alignment with FAIR data principles, Services,
Infrastructure, Knowledge management, Network, Data archiving. We have formulated the
responsibilities, tasks, KSAs and learning objectives per competence area. Our three matrices (one
for each data stewardship role) can be found at https://doi.org/10.5281/zenodo.3239079. Our final
report and all other documents are available at https://zenodo.org/communities/nl‐ds‐pd‐ls/.
We have formulated recommendations related to a) embedding data stewardship roles in university
function profiles, b) developing a self‐assessment tool for the competencies, KSAs and learning
objectives, and c) developing and implementing training. Since September 2019, we are continuing
our work in the context of the National Platform Open Science NPOS
(https://www.openscience.nl/en/projects/project‐f‐education‐and‐training‐in‐open‐science‐and‐
datastewardshop). This new project focuses on professionalizing data stewardship competences and
training. We will build on our previous work as well as on the outcomes of another recent Dutch data
stewardship project from LCRDM (https://doi.org/10.5281/zenodo.2669149).Major partners in the
project are ZonMw, LCRDM, VSNU, Vereniging van Hogescholen, PNN, and SURF.
Acknowledgements This project is funded by the ZonMw Personalised Medicine Programme (Dossier number: 80‐84600‐
98‐3007), Zilveren Kruis en KWF Kankerbestrijding. Additional funding was provided by UMCG,
UMCU, Radboud University Nijmegen, Radboudumc and DTL/ELIXIR‐Netherlands
Keywords: Data stewardship, Research Data Management, Training, Capacity Building, Competences
Skills, FAIR
Abstractbook Health‐RI conference 2020
Abstract: 36 (Poster)
Self‐initiated donation to a biobank. Should and could biobanks offer
this option? E. Vermeulen (1, 2), T. Schaaij‐visser, E. Eijdems (2), S. Rebers (2), M. Kaatee (2), H. Schipper (2), G.
Remmers (2) on behalf of the Maatschappelijke Raad Biobankonderzoek.
(1) VSOP, (2) MAB
Self‐initiated donation to a biobank. Should and could biobanks offer this option?
Introduction: When asked, citizens are very willing to donate tissues to biobanks. Due to increasing
awareness of biobank research and ‘citizen science’ with data tracking, people also consider self‐
initiated donation to biobanks.
Normally, most biobanks do not inform the public about the option of self‐initiated donation. By
providing (international) examples and starting a discussion, the Patient & Public Advisory Council
(PPAC) for Biobank Research (installed by BBMRI.nl) would like to stimulate Dutch biobanks to offer
(information and a procedure about) this option to the public.
Materials and methods: This work was initiated by questions posed by rare disease patient
representatives in the PPAC. Their question resonated with other patient organisations, such as
MD|OG (http://mdog.nl/). A horizon scan was done to collect information and possible procedures,
and to make an inventory of Dutch and other biobanks that welcome self‐initiated donation.
Results: Only a few Dutch biobanks facilitate self‐initiated donation. Self‐initiated donation is possible
under specific circumstances. Some biobanks offer information to donate for example brain tissue
post mortem (https://www.mscnn.nl/doneren/ ).
Some biobanks offer the option for donors to donate while the donor is alive. The International
Fibrodysplasia Ossificans Progressiva association offers people to donate: (
https://www.ifopa.org/biobank ). The Luxemburg biobank (IBBL) offers the option for ‘healthy
controls’ to donate: https://www.biobank.lu/research‐programmes/general‐population/?lang=en.
Tools that can be used to inform citizens are Orphanet and RD‐connect.
Discussion:We conclude there are several, yet still limited options to offer information and
procedures for self‐initiated donations.
Advantages of self‐initiated donation are:
‐ it increases inclusion of citizens‐ advertisement of the work of the biobank
‐ increased engagement of citizens in biobank research
To further discuss the possibilities and limitations, we like to invite all biobanks to visit our poster and
provide their feedback.
Acknowledgements We like to thank BBMRI.NL
Keywords: Public, patients, donation, engagement
Abstractbook Health‐RI conference 2020
Abstract: 37 (Poster)
CBS Microdata Services Fatima El Messlaki (1), Anouk de Rijk (1)
(1) CBS
Statistics Netherlands (CBS) collects data from people, companies and institutions. Upon receipt of
these data, all directly identifying personal details are removed as soon as possible and replaced by a
pseudo key. CBS uses these so‐called pseudonymised data to conduct statistical research. CBS will
never supply identifiable data to third parties. However, (academic) research institutions may, under
strict conditions, be given access to pseudonymised microdata. Microdata are linkable data at the
level of individuals, companies and addresses.
CBS offers a wide range of administrative health care data that can be reused for microdata research.
Combined with surveys and other available data sources with demographic and socio‐economic
variables this offers numerous possibilities for research on health, life style and use of care.
CBS Microdata Services facilitates this use of microdata by employees of authorized research
institutions by giving them access to the data from a secure workplace via a secure internet
connection. The requirements for obtaining an authorization are that the institute’s primary
objective is doing statistical or scientific research and that results of their research are accessible for
the general public. Researchers get access to the datasets necessary to answer their research
question, do their analyses and save their results. Under specified conditions it may also be possible
to link your own microdata set on persons or companies.
Within the RA, research / applications can be done on health care economy, epidemiology and public
health, evaluation or diagnostic and treatment protocols
Via the ODISSEI Microdata Access Discount, employees of ODISSEI can get a discount of up to 50% on
their CBS Microdata projects. ODISSEI also supports the development of the Odissei Secure Super
computer where researchers can analyse their data linked to CBS Microdata in SURFsara’s high‐
performance computing environment.
Acknowledgements not applicable
Keywords: Microdata reused data healthcare registers surveys Super computer CBS Statistics
Netherlands
Abstractbook Health‐RI conference 2020
Abstract: 38 (Poster)
Ethics review for non‐invasive (nWMO) health research: moving
towards a shared approach Martin Boeckhout (1), Miriam Beusink (2), Susanne Rebers (2), Evert‐Ben van Veen (1), Lex Bouter
(3), Irith Kist (2), Marjanka K. Schmidt (2, 4)
(1) MLC Foundation, (2) NKI‐AVL, (3) Free University Amsterdam & AUMC‐Vumc, (4) LUMC
The Wet Medisch‐wetenschappelijk onderzoek met mensen (WMO) provides the regulatory
framework for invasive health research, mandating ethics review for all research falling under its
remit. However, non‐invasive and data‐driven health research generally fall outside the scope.
Upholding standards of research research quality, privacy and data protection measures, as well as
protection of research and data subjects’ rights and interests are just as important for so‐called
nWMO research. However, an overarching national framework for ethics review is currently lacking.
What kinds of research are included under the heading of nWMO research? How is ethics review for
such research currently organized and conducted? What issues are currently at stake, and how can
ethics review be improved upon? Using a mixed‐methods approach involving literature review,
ethical and legal analysis, a series of interviews with ethics committees and stakeholders as well as a
workshop, we set out to answer these questions for an exploratory report commissioned by the
Dutch Ministry of Health. This poster presentation will present the initial findings.
All main parties involved in Dutch health research ascribe to the importance of ethics review for
responsible, high‐quality health research. Ethics review is common, but suffers from high inter‐
institional diversity and fragmentation. Potential ways forward include the drawing up of guidelines
for privacy and data protection, clarification of the burden research can demand from participants
outside the WMO as well as shared risk classifications on which to base organizational policies for
efficient and broadly comparable organization, procedures and formats for ethics review. Broadening
of the scope of the WMO to also include non‐invasive and data‐driven health research received much
support, but will require considerable time and preparation. In the meantime, concerted
collaboration and self‐regulation involving all parties involved in health research provide the most
promising way forward.
Acknowledgements This poster will present the findings of an exploratory report commissioned by the Dutch Ministry of
Health.
Keywords: ethics review, ELSI, privacy, WMO
Abstractbook Health‐RI conference 2020
Abstract: 39 (Poster)
Metadata matters Evelien van der Schaaf‐de Wolf (1), Erik van Iperen (1), Joost Daams (1), Rudy Scholte (1), Silvia
Olabarriaga (1)
(1) Amsterdam UMC, location AMC
When a research project is started, funders and institutions request a Data Management Plan (DMP).
In such DMP, researchers are asked to indicate which metadata standard they will use in order to
enhance data availability for reuse. However, so far no metadata standard has been widely accepted
and adopted for medical data. The DataCite metadata schema is available, however it does not cover
specific information for medical research. The Amsterdam UMC is therefore creating a metadata
schema for medical data.
To comply with standards that already exist, our first step was to select items from the DataCite
metadata schema that were considered meaningful for medical research. The second step was to
identify domains that use similar data types within the Amsterdam UMC. For each domain, an expert
was asked to determine a minimal set of metadata.
This resulted in two main metadata categories: the data collection level and the domain‐specific
level. The data collection level consists of a subset of DataCite items supplemented with additional
items that are necessary as a useful description for medical data collections. The domain‐specific
level consists of items for the following: subject data, biosamples, genetic data, images and signals,
and questionnaires. To facilitate initial adoption it is not mandatory to use all items, but it is
recommended to indicate as many as possible.
In the near future the metadata schema will be tested at the Amsterdam UMC through pilot projects.
Researchers who have started their project about four years ago will be asked to use the metadata
schema to describe their study data. In the long term, the vision is to implement the metadata
schema in a research data management tool (e.g. iRODS) in which all Amsterdam UMC research data
can be archived.
Acknowledgements We thank the contribution of the domain experts, Paul Groot and Aldo Jongejan, and the feedback
received from the Data Matters Expert Group of the Amsterdam UMC.
Keywords: metadata schema, data management plan (DMP), archiving, medical data
Abstractbook Health‐RI conference 2020
Abstract: 40 (Poster)
Getting started with trusted FAIR data lakes Erik Flikkenschild (1), Marlon Domingus (2)
(1) LUMC, chairman NL‐SIG veilige datakoppelingen, (2) Erasmus University Rotterdam, Data
Protection Officer
Introduction
Building trustworthy multipurpose data lakes requires a multi functional (legal, ethical and technical)
approach. Pooling pseudonomised data is complex, time consuming and experts must deal with
substantial privacy risks. Legal discussions are regretfully not started from community accepted
ethical viewpoints, and in the discourse, different perspectives are not fully done justice by not
recognizing their different scopes. Expert IT working groups, for example, typically do not have a
common accepted legal basis to start with, and tend to solve domain specific challenges, thus
creating silos.
Strategy
The authors of this paper are convinced that progress will be made if we start with a community
accepted ethical viewpoint first, in which common interest and privacy concerns are balanced. The
validity of these ethical principles shall subsequently be validated in the GDPR norms and formulated
in a common interest NL GDPR manifest. IT architects can then add technical safeguards in order to
secure the required trust. Everyone has to take into account the different perspectives (Open
Science, Artificial Intelligence, Personalized Medicine,..)
Methods
Datalake design principles (trusted linkage data framework) to be discussed based on the results of
ethical guiding principles. With a POC one can take a first small step and create an IT architecture for
each perspective. Evaluation of these three designs by existing NL communities can validate this
collaborative approach and decide on suitable adequate measures.
Results
A (national) trusted linked data framework and method that can be used to build trustworthy data
lakes.
Follow‐up
The creation of a national multi‐disciplinary multi‐domain working group that specifies the trusted
linked FAIR data framework.
Acknowledgements LCRDM working groups, SIG Veilige koppelingen members, NFU Good Research Practice working
group 6
Keywords: data infrastructures, Trusted data datalakes, ethics, personal data linkage, Open Science, Personalized Medicine, Artificial Intelligence, AVG, GDPR, data driven, ELSI
Abstractbook Health‐RI conference 2020
Abstract: 41 (Poster)
The real world nature of Prospective Dutch ColoRectal Cancer cohort
(PLCRC) Jeroen W.G. Derksen (1), Merel Wassenaar (1), Marloes A.G. Elferink (2), Jeanine M.L. Roodhart
(1), Anne M. May (3), Miriam Koopman (1), Geraldine R. Vink (1,2), on behalf of the PLCRC Working
Group
(1) Department of Medical Oncology, University Medical Center Utrecht, Utrecht, Netherlands, (2)
Department of Research, Netherlands Comprehensive Cancer Organisation, Utrecht, Netherlands, (3)
Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht,
Netherlands
Background: Large high‐quality population‐based cohort studies are of tremendous value to support
real‐world data studies and improve treatment outcomes. In 2013, the Prospective Dutch ColoRectal
Cancer cohort (PLCRC) was initiated combining different data sources collecting longitudinal clinical
data, patient‐reported outcomes and biomaterial. PLCRC serves as an infrastructure for a wide range
of observational and interventional studies to improve outcomes of colorectal cancer (CRC) patients.
Here, we investigate whether PLCRC evolves in the direction of a nationwide cohort of real world
nature.
Methods: All CRC patients 18 years and older are eligible for PLCRC. Clinical and demographical data
of PLCRC participants, as collected in the Netherlands Cancer Registry, are compared with the total
Dutch CRC population over the period 2013‐2017 (reference population), also obtained from the
Netherlands Cancer Registry. Cohort characteristics are described and populations compared.
Results: In August 2019, 5722 patients were enrolled, of which 4759 (83%) with a complete TNM
stage classification included in the analyses. Compared to participants enrolled between 2013‐2016
(N = 1,088), we found a small shift of the 2017‐2019 population (N = 3671) towards the Dutch
reference population (N=72.685) in terms of age at diagnosis (mean 64.6±10.2 years in 2013‐16,
65.0±10.2 in 2017‐19, and 69.3±10.8 in the ref. group), sex (65% males in 2013‐16, 61% in 2017‐19,
and 57% in the ref. population), location of primary tumor (56% rectum in 2013‐16, 39% in 2017‐19,
and 31% in the ref. population) and TNM stage (34% stage I‐II in 2013‐16, 42% in 2017‐19, and 49% in
the ref. population).
Conclusion: Over the past years, enrollment in PLCRC steeply increased. Improvements in
recruitment and multidisciplinary enrolment of patients has enhanced PLCRC’s representation of the
real‐world. This helps to learn from today’s patients to enable personalized therapy facilitating better
outcomes for future CRC patients.
Acknowledgements n.a.
Keywords: infrastructure, cohort, real‐world data, personalized treatment, data collection
Abstractbook Health‐RI conference 2020
Abstract: 42 (Poster)
HOVON Pathology Facility and Biobank: Making the right choices for
workflow and data Nathalie J. Hijmering (1, 2), Erik van Iperen (3), Phylicia Stathi (1, 2), Dirk Veldman (4), Paula
Rinkens (4), Karin Aretz (4), Rita Azevedo (5), Martine Chamuleau (6), King H. Lam (7), D. De Jong
(1)
(1) Department of Pathology, AmsterdamUMC, location VUmc, Amsterdam, The Netherlands, (2)
HOVON Pathology Facility and Biobank, AmsterdamUMC, location VUmc, Amsterdam, The
Netherlands, (3) Durrer Center for cardiovascular research, Netherlands Heart Institute, Utrecht, The
Netherlands, (4) MEMIC, Maastricht University, Maastricht, The Netherlands, (5) Lygature, Utrecht,
The Netherlands, (6) Department of Hematology, AmsterdamUMC, location VUmc, Amsterdam, The
Netherlands, (7) Department of Pathology, Erasmus University Medical Center, Rotterdam, The
Netherlands
Background
Central pathology review and translational studies on tissue biopsy material are an integral part of
clinical trials for malignant lymphoma patients performed by HOVON (Haemato‐Oncology
Foundation for Adults in the Netherlands). The HOVON Pathology Facility and Biobank (HOP)
supports all pathology‐ related activities from requesting and processing the pathology material until
support of side studies. Optimized logistics are required to improve the quality and speed of
pathology review and successfully accommodate translational research.
Methods and Result
We have developed a customized, web‐based platform to monitor requesting, dispatching, collecting
and processing of bio‐specimens and storage of all compiled biomarker data within the TraIT
(Translational Research IT) infrastructure. Four years of experience now shows that complete
pathology review results can be made available within weeks after completion of the trial, including
molecular information. However, the current system is limited by its frozen design after launch,
restricted options for integration with external tools (double data‐entry) and suboptimal data export
options. Therefore, we designed a future‐proof platform, based on our experience. We separated the
logistical platform (Ldot) from the data storage (Castor), thereby introducing flexibility. Both
platforms are highly suitable to be set up by the user with support from MEMIC and Castor. Ldot
provides a user‐friendly overview of ongoing actions supporting the daily workflow. Both platforms
are optimized for integration to receive and export data from/to external tools, such as ALEA,
EXCEL/SPSS.
Conclusions
Future‐proof workflow platforms such as the HOP benefit from a flexible, modular design that can be
fully set‐up and maintained by the user. Various tools within the design should be selected for user‐
friendly daily workflow support and integrative options with external systems for data import and
export to optimize performance for direct trial‐related actions as well as for future research
applications according to FAIR principles.
Acknowledgements No
Keywords: Workflow support, Clinical trials, Pathology, Lymphoma, TraIT, Ldot, Castor
Abstractbook Health‐RI conference 2020
Abstract: 43 (Poster)
Public radiomics data collections in an open access Semantic Web
(SPARQL) endpoint Petros Kalendralis (1), Zhenwei Shi (1), Chong Zhang (1), Ananya Choudhury (1), Alberto Traverso
(1), Matthijs Sloep (1), Johan Van Soest (1), Rianne Fijten (1), Andre Dekker (1), Leonard Wee (1)
(1) GROW School for Oncology and Developmental Biology‐ Maastricht University Medical Centre+,
Department of Radiation Oncology MAASTRO, Maastricht, The Netherlands
Purpose or Objective
In a groundbreaking investigation, Aerts et al. (1) showed that quantitative imaging features
(radiomics) could potentially be used to decode information about tumour phenotype that is
relevant to disease prognosis. This publication has been the subject of intense interest ever since,
and there have been numerous requests for more information about the datasets – RIDER,
Interobserver, Lung1 and Head‐Neck1. To support research into repeatability, reproducibility,
generalizability and explainability in radiomics, we have now made the clinical follow‐up, extracted
pyRadiomics (2) features and DICOM metadata findable, accessible, interoperable and re‐usable
(FAIR) through a public semantic web access point (http://sparql.cancerdata.org).
Material and Methods
Overall survival intervals (days since start of radiotherapy) have been updated through the Dutch
national registry under an internal ethics board‐approved request. Spatially incorrect offsets of the
Primary Gross Tumour Volume (“GTV‐1”) regions of interest (ROIs) in the Lung1 set were amended in
The Cancer Imaging Archive (TCIA) collection . Image features were extracted using the ontology‐
guided radiomics workflow (3) (O‐RAW) and published in Resource Descriptor Format (RDF)
consistent with the Image Biomarker Standardization Initiaitive (IBSI) through an open Radiomics
Ontology. DICOM metadata as RDF was extracted using a research version of Semantic DICOM
(SoHard, GmbH, Fuerth; Germany). Clinical data was published in RDF using the Radiation Oncology
Ontology. Example queries were tested, which verified that the SPARQL endpoint was accessible.
Conclusion
We successfully generated separate RDF repositories of clinical, DICOM and radiomics data and
published these on an open access SPARQL endpoint. We can effortlessly cross‐reference into the
clinical, dicom and radiomics data. Queries can be generated which simultaneously looks in all three
repositories, thus taking advantage of the semantic linking between the data elements.
References
1: PMID: 24892406
2: PMID: 29092951
3:PMID: 31580484
Acknowledgements Clinical Data Science group‐Maastricht University Medical Centre+, Department of Radiation
Oncology MAASTRO, Maastricht, The Netherlands.
Keywords: Keywords: Radiomics, public datasets, reproducibility, FAIR data
Abstractbook Health‐RI conference 2020
Abstract: 44 (Poster)
The FAIRification of clinical data with modular knowlegde graphs. Matthijs Sloep (0000‐0003‐3602‐1885) (1), Petros Kalendralis (1), Johan van Soest (0000‐0003‐
2548‐0330) (1), Rianne Fijten (0000‐0002‐1964‐6317) (1), Andre Dekker(0000‐0002‐0422‐7996) (1)
(1) Department of Radiation Oncology (MAASTRO), GROW school for Oncology and Developmental
Biology, Maastricht University Medical Centre+, Maastricht, The Netherlands
Everyday physicians manually collect and enter information about their patients in electronic records
and turn it into some of the most expensive data available. Combining data from multiple centres is
vital for good research, unfortunately, linking it is not straightforward due to differences in treatment
protocols and clinical systems. Consequently, before we can solve any medical and research
problems, we first need to solve data integration problems first. ProTRAIT is an effort to link patient
data from multiple radiotherapy centres to build a comprehensive national registry and research
database for proton beam therapy. We turned to the FAIR principles
(https://doi.org/10.1038/sdata.2016.18) to facilitate data integration. We address the FAIR principles
using a Semantic Web technology and describe the challenges we faced and our solutions to solve
these issues.
Our approach to the FAIRification process is akin to the steps described by the Go‐FAIR initiative.
Radiotherapists first listed specific clinical items, then we manually created a knowledge graph and
annotated it with existing and new ontology classes using the Radiation Oncology Ontology
(https://doi.org/10.1002/mp.12879). Our approach means that much effort goes into creating and
maintaining the knowledge graphs. Each cancer type requires a separate, manually created
knowledge graph. To facilitate this process, we made full use of the considerable overlap between
information elements by creating separate turtle files for small subsets of information elements to
leverage the modular characteristics of knowledge graphs. The challenge was to arrange these items
in subsets that are both logical and practical, but the big advantage of this modular approach is that
we can easily adapt and add more variables when needed to adjust our graphs to changes in the
clinic.
Acknowledgements KWF Kankerbestrijding
Keywords: FAIR data, Linked Data, Modular, knowledge graphs
Abstractbook Health‐RI conference 2020
Abstract: 45 (Poster)
Privacy Sensitive Distributed Analysis of Dementia Cohorts from
Hospitals in The Netherlands Stuti Nayak (1), Ananya Choudhury (1), Matthijs Sloep (1), Inigo Bermejo (1), Johan Van Soest (1),
Andre Dekker (1)
(1) Department of Radiation Oncology (MAASTRO), GROW school for Oncology and Developmental
Biology, Maastricht University Medical Centre+, Maastricht, The Netherlands
Dementia is a multi‐factorial disease that affects around 35 million people around the world. The
prediction is that in Netherlands the absolute number of patients will increase by 115% in the next 20
years. (volksgezondheidenzorg.info) It imposes an enormous burden on society with both the
suffering of patients and their caregivers and the tremendous financial costs.
Although there is ample data around, not all data can be used to predict better outcomes for
patients. Not only technical bottlenecks but also ethical, legal and societal issues impede data sharing
and hinder researchers from using data to its full potential. Also, there is minimal standardization of
data and as such, often data from different hospitals are syntactically and semantically not
interoperable with each other.
However, with Personal Health Train (PHT), the focus shifts from sharing data to sharing algorithms
data. The PHT is agnostic of the actual data and relies heavily on the Findable Accessible
Interoperable and Reusable data principles. The current project aims at establishing a FAIR data‐
sharing infrastructure, the Personal Health Train, which connects three hospitals in the Netherlands
namely, EMC, Rotterdam, MUMC, Maastricht and LUMC, Leiden. The infrastructure will enable
usability of sensitive dementia cohorts from each of these hospitals, without data having to leave the
hospital.
As a proof of concept, we set up two mock FAIR data stations containing minimal set of variables
such as age, gender, diagnosis, MMSE, smoking, cholesterol, diabetes, BMI and hypertension. We
designed the Dementia Cohort Analysis (DCA) train to analyze the distributed cohort and find
correlations between the variables. It has been shown that we can use sensitive data from multiple
sources using PHT and still adhere to the ELSI of data sharing. These correlations would further lead
to designing and training the machine learning models for risk prediction in dementia patients.
Acknowledgements MEMORABEL Project, Department of Radiation Oncology (MAASTRO)
Keywords: Personal Health Train, FAIR, distributed learning, dementia
Abstractbook Health‐RI conference 2020
Abstract: 46 (Poster)
Only one copy H. Pieterman (1)
(1) ErasmusMC Rotterdam
Only one copy.
To do their job safely, physicians need access to all medical information of their patients. In fact,
diagnosis and treatment should be based on interpretation of both actual and all historical
information. It also should be possible to relate all these different information data.
All data should therefore ideally be displayed along one timeline. However the medical history of a
patient is not presented in the form of chronological data, but the data are traditionally more or less
hidden in documents in the various patients’ files, each with its own timeline. Nowadays, nearly 25%
of patients visit more than one physician (general practitioner not included) in parallel, and in
different hospitals, causing both fragmentation and duplication of patient files. A solution could be a
national electronic patient file. However, until today this is politically unacceptable in the
Netherlands. As a result, much time and money is spent in developing a platform solution for
searching, viewing and eventually copying the medical data.
A better solution which prevents searching, is storing all patient information (in the format of data)
immediately after their generation in a personal database hosted in one of multiple dedicated
datacenters. The data in such a “health vault” can only be accessed by the patients themselves or by
those who have an actual treatment relationship with the patient. Because the general practitioner
has a lifetime treatment relation with his patient he could serve as a steward or keyholder.
In doing so, data will be findable, accessible, interoperable and re‐usable (FAIR) for healthcare. With
commitment of the patients these datacenters can also function as stations for personal health
trains (PHT) for research.
Acknowledgements none
Keywords: timeline, patient files, database, FAIR, PHT
Abstractbook Health‐RI conference 2020
Abstract: 47 (Poster)
Personal Health Train Coalition I.M. Tharun (1)
(1) Lygature
Acknowledgements
Keywords:
Abstractbook Health‐RI conference 2020
Abstract: 48 (Poster)
Implementation and Deployment of a Federated Logistic Regression Arturo Moncada‐Torres (1), Frank Martin (1), Katja Aben (1),
Stefan Willems (2), Rinus Voorham (2), Paul Seegers (2), Gijs Geleijnse (1)
(1) Netherlands Comprehensive Cancer Organization (IKNL), Eindhoven, NL, (2) PALGA Foundation,
Houten, NL
At the Netherlands Comprehensive Cancer Organization (IKNL), we are dedicated to continuously
improve care for cancer patients. For this purpose, we curate the Netherlands Cancer Registry, a
population‐wide registry containing data of nearly all cancer cases in the country since 1989.
Complementing IKNL’s clinical data with (molecular) pathological data from the National Registry of
Histo‐ and Cytopathology (PALGA) has been proven to be of unprecedented value for observational
research. This is evidenced by the more than 30 joint projects between these two institutions
annually.
For these projects, data are typically obtained by aligning both datasets and generating a centralized
copy of the data, which can then be used for analysis. However, the latter is not ideal, since it could
potentially compromise patient data privacy. As a matter of fact, the General Data Protection
Regulation (GDPR) has established strict rules and limitations in this matter. Moreover, centralizing
data poses several organizational, operational, social, and political challenges.
In this project, we used our open‐source priVAcy preserviNg federaTed leArninG infrastructurE
(VANTAGE) for jointly analyzing IKNL and PALGA data without them leaving their original location.
Namely, we implemented a federated logistic regression model (based on the work by Li et al., 2015)
to investigate the relation between structured pathological reporting and survival of prostate cancer
patients. The federated model’s coefficients were equivalent to their centralized counterparts. The
results of this project show the potential for Federated Learning using VANTAGE as a cornerstone of
the Personal Health Train.
Acknowledgements ‐
Keywords: distributed learning, federated learning, personal health train, vertically‐partitioned data
Abstractbook Health‐RI conference 2020
Abstract: 49 (Poster)
Secure Log Rank Test in Survival Analysis on Vertically Partitioned Data
using Multi‐Party Computation Rooijakkers, T.A. (Thomas) (1), Kamphorst, B. (Bart) (1), l’Isle, N.A.F. (Natasja) van de (1),
Cellamare, M. (Matteo) (2), Knoors, D. (Daan) (2)
(1) TNO, (2) IKNL
The growing complexity of cancer diagnosis and treatment requires data sets that are larger and
richer than currently available in a single cancer registry. However, sharing patient data is difficult
due to patient privacy and data protection needs. Secure Multi‐Party Computation (MPC) has the
potential to overcome these limitations. MPC is an umbrella term for cryptographic techniques that
allows several different entities to jointly perform analysis on data without sharing their actual data.
IKNL and TNO are collaborating to develop solutions using these technologies to enable privacy‐
preserving training of survival analysis models (e.g. Kaplan‐Meier estimator, Log Rank Test, Cox
regression, etc.)
The Kaplan–Meier estimator is a non‐parametric statistic used to estimate the survival function of a
lifetime table. To compare survival between groups we can use the log rank test. The log rank test is
a statistical procedure that compares two (or more) survival distributions by testing, at each
observed event time, whether the hazard functions of the groups are different. A direct application
could be to test whether treatment X has a greater effect on the longevity of a patient compared to
another treatment Y.
We present an MPC protocol that permits to perform the log rank test on vertically partitioned data.
In particular, we focus on cases where, for a group of patients, party A owns data on the patients
survival (diagnosis date, death date, censorship, etc.) and party B has access to the treatment
information.
Acknowledgements 1. Veugen, P.J.M. (Thijs)
Keywords: Multi‐Party Computation, MPC, Survival Analysis, Kaplan‐Meier, Log Rank Test
Abstractbook Health‐RI conference 2020
Abstract: 50 (Poster)
Characterization of depression symptoms using large scale
questionnaire data in the Dutch population: a BBMRI‐BIONIC study Marije van Haeringen (1)#, Sarah R Vreijling (1)#, Floris Huider (2), Mariska Bot (1), Yuri Milaneschi
(1), Najaf Amin (3), Joline W. Beulens (4, 5, 6), Marijke A. Bremmer (1), Petra J. Elders (5, 7), Tessel
E. Galesloot (8), Lambertus A. Kiemeney (8), Hanna M. van Loo (9), H. Susan J. Picavet (10), Femke
Rutters (4, 5), Ashley van der Spek (3), Anne M. van de Wiel (11), Cornelia van Duijn (3, 13), Edith
J.M. Feskens (11), Catharina A. Hartman (9), Albertine J. Oldehinkel (9), Jan H. Smit (1), W. M.
Monique Verschuren (6, 10), Gonneke Willemsen (2), Eco JC de Geus (2), Brenda WJH Penninx (1),
Dorret I Boomsma (2), Femke Lamers (1)*, Rick Jansen (1)* # shared first author, *shared last
author
(1) Department of Psychiatry, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam Public
Health Research Institute and Amsterdam Neuroscience, Amsterdam, Netherlands, (2) Department of
Biological Psychology, Vrije Universiteit Amsterdam, Amsterdam, Netherlands, (3) Department of
Epidemiology, Erasmus University Medical Center, Rotterdam, The Netherlands, (4) Department of
Epidemiology and Biostatistics, Amsterdam University Medical Centres, location VUMC, The
Netherlands, (5) Amsterdam Public Health Research Institute, The Netherlands, (6) Julius Center for
Health Sciences and Primary Care, University Medical Center Utrecht, University Utrecht, Utrecht, The
Netherlands, (7) Department of General Practice, Amsterdam University Medical Centres, The
Netherlands, (8) Radboud university medical center, Radboud Institute for Health Sciences, Nijmegen,
The Netherlands, (9) Department of Psychiatry, University of Groningen, University Medical Center
Groningen, Groningen, Netherlands, (10) Centre for Nutrition, Prevention and Health Services,
National Institute for Public Health and the Environment, Bilthoven, The Netherlands, (11) Division of
Human Nutrition and Health, Wageningen University, Wageningen, The Netherlands, (12)
Amsterdam Public Health and Amsterdam Neuroscience, The Netherlands, (13) Nuffield Department
of Population Health, University of Oxford, Oxford, UK
Background
Depression is a highly heterogeneous disease with diverse symptom profiles. In current clinical
practice, a personalized approach based on symptoms or biomarkers is lacking. The BIObanks
Netherlands Internet Collaboration (BIONIC) within the BBMRI infrastructure is a large scale online
survey on lifetime depression in seven Dutch population‐based and clinical cohorts. Our aim here is
to explore the consistency of single symptom prevalence across cohorts and determine demographic,
clinical and lifestyle characteristics of single symptoms.
Methods
Data was obtained from seven cohorts (N=~75.000, 60% female, age 16‐100) with the online Lifetime
Depression Assessment Self‐report (LIDAS). Lifetime depression was defined according to DSM‐5
criteria. The LIDAS contains the 8 DSM symptoms for depression and additionally demographic and
physical characteristics, information on age of onset, number and duration of episodes, past
treatment, and comorbidity with other psychiatric conditions. Using linear or logistic models, for
each of the 8 symptoms we compared participants with and without a specific depression symptom
in these demographic, clinical and lifestyle characteristics.
Results
Abstractbook Health‐RI conference 2020
Preliminary analyses based on ~ 4.000 individuals with lifetime depression revealed that, besides the
highly prevalent core DSM symptoms ‘depressed mood’ and ‘loss of interest’, the symptoms ‘trouble
concentrating’ and ‘energy loss’ had the highest prevalence (>80%). Cohorts showed very similar
prevalence rates of depression symptoms. The symptom ‘increased appetite or weight’ (prevalence:
20%) appeared to have the most distinctive demographic and clinical profile as compared to other
symptoms: individuals with this symptom were younger (P<1e‐7), more often female (P<1e‐7), had
more often recurrent episodes (P<1e‐17), and were younger during their first episode (P<1e‐8).
Conclusion
We found support for the consistency in endorsement of individual depressive symptoms across
cohorts. Patients with increased appetite or weight are most different from other patients, which
may indicate a partially unique underlying pathophysiology. We will use this information in our
future research to investigate the role of individual depression symptoms in personalized medicine
approaches.
Acknowledgements This research was financially supported by BBMRI‐NL,
a Research Infrastructure financed by the Dutch
government (NWO 184.021.007). We would like to
acknowledge all researchers involved in the BBMRI‐NL
project ’Phenomics 2.0 ‐ proof of principle for major depressive disorder’
Keywords: Online survey, depression
Abstractbook Health‐RI conference 2020
Abstract: 51 (Poster)
What is a “digital” patient? An ontological approach. L.P. Ter Meer (1)
(1) Erasmus School of Health Policy and Management, Rotterdam
There is a need to define essential objects so we can describe role, function and purpose and be able
to register these objects in a uniform way. “Patient” is one of the most frequently used objects in
healthcare but it lacks a uniform ontological description. Patient is most often used as a subject of a
disease instead of having its own identity and related elements. We describe an ontological model
for a patient encompassing 2 classes of elements, namely patient objective and patient subjective
ones. The model when applied may assist researchers, care providers, software developers and users
of the terminology in a more consistent approach of the patient. The advantage for the patient is
that the model offers an easy and complete overview of the components influencing its health and
when values are out of range how they can be related to disease. The model may also facilitate in the
design and process of data exchange.
Acknowledgements Dr. M. de Mul; Dr. E. Veringa
Keywords: Patient; Electronic medical record; data exchange; ontological definition; patient model
Abstractbook Health‐RI conference 2020
Abstract: 52 (Poster)
FAIRification at DNA, RNA and protein level in studying colorectal
tumor progression Menno de Vries (1), Malgorzata A. Komor (1), Mariska Bierkens (1), Annemieke C. Hiemstra (1),
Stefan Parayble (2), Wibo Pipping (2), Jan Hudecek (1), Guido Jenster (3), Gerrit A. Meijer (1),
Remond J.A. Fijneman (1)
(1) Department of Pathology, The Netherlands Cancer Institute, Amsterdam, (2) The Hyve, Utrecht, (3)
Department of Urology, Erasmus Medical Center Rotterdam
Colorectal adenomas, carcinomas and normal adjacent colorectal tissues were characterized at the
DNA, RNA and protein level as part of the NGS‐ProToCol study, in order to get a better understanding
of colorectal tumor progression. Large amounts of data were generated during this project, and
efforts were made to work towards FAIRification (Findable, Accessible, Inter‐operable and Re‐usable)
of the different data types; from raw data to processed or ‘final’ data, as well as accompanying
metadata.
Several applications were used to accommodate the FAIRification of diverse data types collected
during the study. The raw sequencing files (FASTQ) and accompanying phenotypic metadata were
uploaded to the European Genome‐phenome Archive (EGA); digital images of Tissue Microarrays
(TMA) were uploaded to the SlideScore server, hosted by the NKI; and all processed or ‘final’ data
(clinical, pathology, biosample and molecular data) were imported to the national instance of
tranSMART, a data‐integration platform specialized in cohort‐centric data selection/exploration, and
partly to cBioPortal. In tranSMART, metadata and hyperlinks were used to link to the raw data in
EGA, and to individual TMA cores in SlideScore. As a result, an overview of all existing data is
available in a user‐friendly way. In the nearby future, the full study will also be uploaded to
cBioPortal, another data‐integration platform hosted by Health‐RI. It has complementary view and
query capabilities, being specialized in sample, longitudinal and gene‐centric views and queries
combined with clear visualizations.
In conclusion, the rich dataset of NGS‐ProToCol can be re‐used for scientific research, resulting from
FAIRification of both its raw and processed data.
References
‐ European Genome Archive, https://www.ebi.ac.uk/ega/home
‐ SlideScore, www.slidescore.com
‐ tranSMART, https://trait.health‐ri.nl/trait‐tools/transmart
Acknowledgements All people who were involved with NGS‐ProToCol
Keywords: Colon, Carcinoma, EGA, European Genome Archive, tranSMART, SlideScore, NGS‐ProToCol
Abstractbook Health‐RI conference 2020
Abstract: 53 (Poster)
The Handbook for Adequate Natural Data Stewardship (HANDS) Els Swennen (1), Pascal Suppers (1), Tom Delnoy (1), Petra van Overveld (2), Marco Roos (2), Sonja
Meeuwsen (2), Salome Scholtens (3), Bert van Ooijen (4), Chantal Steegers (5), Klaudia Onnasch (5),
Erik van Ieperen (5), Rudy Scholte (5), Jeroen Belien (5), Sander de Ridder (5), Ronald van Schijndel
(5), Gepke Uiterdijk (5), Susanne Rebers (6), Ingeborg Verheul (7), Jan‐Willem Boiten (8), Linda van
den Berg (9) and Paula Jansen (10)
(1) Maastricht UMC+, (2) LUMC, 3. UMCG, (4) Erasmus MC, (5) Amsterdam UMC, (6) ELSI
Servicedesk/NKI, (7) LCRDM, (8) D4LS/Lygature, (9) WASHOE Life Science Communications, (10) UMC
Utrecht
Data stewardship refers to sustainable care for research data as integral part of the research process.
It covers all actions necessary to make digital research data Findable, Accessible, Interoperable and
Reusable (FAIR) during and after your research project, including data management, archiving and
reuse by third parties.
The Handbook for Adequate Natural Data Stewardship (HANDS) provides researchers at the eight
Dutch University Medical Centres (UMCs) with guidelines on data stewardship as well as lists of
practical steps to take for each stage of the research data life cycle. HANDS is one of the services and
tools that is offered on the Health‐RI website. A toolbox in HANDS refers to additional expertise and
resources within or outside your UMC. It offers information for all people involved in data
stewardship, from researchers and data stewards, to policy makers and developers of IT
infrastructure.
Acknowledgements Bijdrage 2015 versie: Peter Doorn (DANS, KNAW), Rob Hooft (DTL), Evert van Leeuwen
(Radboudumc), Leendert Looijenga (Federa), Barend Mons (DTL, LUMC), Arnoud van der Maas
(Radboudumc), Ronald Brand (LUMC), Morris Swertz (UMCG), Jan Jurjen Uitterdijk (UMCG), Pieter
Neerincx (UMCG), Jan Hazelzet (Erasmus MC), Linda Mook (Erasmus MC), Thijs Spigt (TTO, Erasmus
MC), Evert Ben van Veen (MedLawConsult), Margreet Bloemers (ZonMw), Jan Willem Boiten (CTMM‐
TraIT), Cor Oosterwijk (VSOP), Tessa van der Valk (VSOP), Jaap Verweij (Erasmus MC).
Keywords: Guidelines, datastewardship, FAIR, Toolbox
Abstractbook Health‐RI conference 2020
Abstract: 54 (Poster)
SURF Research Access Management, an authorisation and
authentication service optimised for researchers Rogier de Jong (1), Raoul Teeuwen (1), Michiel Schok (1)
(1) https://www.surf.nl/en/pilot‐authentication‐en‐autorisation‐for‐research‐services
SURF is a cooperative association of Dutch educational and research institutions in which its
members join forces. The members are the owners of SURF.
Researchers often experience problems logging in to research services. In order to make logging in
safe, easy and efficient, SURF has been conducting pilots with approximately 10 institutions with an
authorisation and authentication service optimised for researchers: SURF Research Access
Management (SRAM).
The service will be officially launched in Q2 2020 and is also applicable to existing / future Health RI
services.
SRAM tries to solve a number of specific challenges faced by researchers:
‐ How do you arrange consent and logging of access so that you comply with the GDPR?
‐ How can 'guests' from other institutions, companies or outside the Netherlands make use of
research services?
‐ How is group management arranged?
‐ How do we deal with specific research attributes?
‐ How can institutions limit the administrative workload resulting from having to create guest
accounts, 0‐hour contracts, etc.?
‐ How do you arrange access to non‐web resources (such as SSH) based on an institution account?
More info: https://www.surf.nl/en/pilot‐authentication‐en‐autorisation‐for‐research‐services
Acknowledgements SURF
Keywords: Identity & Acces Management for research, Trust and Identity, SURF
Abstractbook Health‐RI conference 2020
Abstract: 55 (Poster)
Administration of research logistics Dirk Veldman (1), Annemie Mordant (1) , Jacqueline Pisters (1), Luc Linden (1), Alfons Schroten(1)
(1) Maastricht University,MEMIC, Centre for data and information management
With large quantities of patients, extensive calling lists, SMS reminders and more, keeping track of a
research project can be challenging.
By using Ldot, you can build your own schedule that fits your research needs and preferences
perfectly. The Ldot Study Builder helps you simplify the execution of daily tasks in large studies
and/or complex protocols.
Benefits of Using Ldot
1. Have an immediate insight into study status pro‐active and visualize progress
2. Standardize protocol execution
3. Minimize required team efforts
4. Secure central storage of logistical data
5. Separate storage of personal data
6. Integration with data collection tools like Castor EDC, Open Clinica, and Qualtrics
Features
Ldot also offers a wide range of different features. These features include:
• GCP compliant (Good Clinical Practice guidelines)
• GDPR compliant (General Data Protection Regulation)
• Secure data storage (ISO 27001 certified)
• Secure and extended user role management
• Compatible for multicentre projects
Acknowledgements TraIT
Keywords: Logistics , GDPR, SelfService, Multicentre
Abstractbook Health‐RI conference 2020
Abstract: 56 (Poster)
Cardio‐metabolic profiling ‐ Association of EFV with increased levels of
circulating lipid metabolites Fariba Ahmadizar (1), Maxime Bos (1), Daniel Bos (1,2), Arfan Ikram (1), Mohsen Ghanbari (1),
Maryam Kavousi (1)
(1) Department of Epidemiology, Erasmus University Medical Center, Rotterdam, The Netherlands ,
(2)Department of Radiology and Nuclear Medicine, Erasmus MC ‐ University Medical Center,
Rotterdam, the Netherlands
Background/Objectives. Recent evidence highlights a link between larger epicardial fat volume (EFV)
and an unfavorable cardio‐metabolic profile. We explored the association of plasma lipid metabolites
with EFV among general population. We also performed the analyses in a subset of subjects with
type 2 diabetes (T2D).
Methods. We included a total of 695 participants from the population‐based Rotterdam Study.
Plasma samples were collected between 2002 and 2005 and metabolites were measured by proton
nuclear magnetic resonance (NMR). The assessment of EFV was through cardiac and extracardiac
multidetector computed tomography (MDCT), quantified in millilitres. Linear regression analysis
adjusted for age, sex, BMI, lipid‐lowering medications and smoking and corrected for multiple testing
(P‐value 0.05/142 independent lipid metabolites = 3.5 × 10‐4) was used to assess cross‐sectional
associations between EFV and 202 lipid metabolites.
Results. After correction for multiple testing, 102 lipid metabolites were independently associated
with EFV; the strongest positive association was shown with phospholipids in large VLDL (beta: 0.1;
SE: 0.01; p‐value: 9.9 × 10‐16), triglycerides, lipids in VLDL subclasses, and apolipoprotein B. Higher
levels of circulating phospholipids in large VLDL were significantly associated with larger EFV in
individuals with T2D (beta: 0.07; SE: 0.03, p‐value:1.6 × 10‐4). In individuals free of diabetes, the lipid
profile was similar to general population, except for phospholipids where the association was not
significant.
Conclusions. Larger EFV was associated with increased levels of circulating lipid metabolites mainly
phospholipids in large VLDL. Phospholipid metabolism has shown a central role in pathogenesis of
metabolic disease and is associated with insulin resistance and T2D. Association of lipidomics
signature to fat deposit may help provide more biological insights into risk stratification for metabolic
outcomes.
Acknowledgements The dedication, commitment, and contribution of inhabitants, general practitioners, and pharmacists
of the Ommoord district to the Rotterdam Study are gratefully acknowledged.
Keywords: Epicardial fat, lipidomics, type 2 diabetes
Abstractbook Health‐RI conference 2020
Abstract: 57 (Poster)
Amsterdam UMC expertise center for high performance computing Bob W. van Dijk (1), Paul F.C. Groot (1), Martijn D. Steenwijk (1), Daoud Sie (1), Ronald van
Schijndel (1)
(1) Amsterdam University Medical Centers, Amsterdam, The Netherlands
A number of trends make it urgent to improve high performance computing (HPC) services at
Amsterdam UMC:
• The extreme growth of data use in health care and research,
• The increasing demand for sharing data and analysis methods,
• The stricter regulations regarding data safety and privacy,
• The trend toward open science and FAIR data management,
• And the growing interest in artificial intelligence and machine learning.
As a result of these developments, between 300 and 400 researchers in Amsterdam UMC demand
better computing facilities. In surveys, these researchers responded a need for more compute
power, more data storage capacity, more flexibility in application use, and better ways to share data
than are available in the standardized ICT work environment. This mismatch between the centralized
ICT facilities and research needs has led to a fragmented landscape of self‐managed local solutions
for HPC that lacks cost efficacy and carries many risks.
To provide tailored HPC solutions, a center of expertise has been set up consisting of an HPC‐facility,
an HPC‐hub and an HPC‐community. The HPC expertise center has a front‐office and is positioned
within Research Support.
Amsterdam UMC aims to provide researchers with adequate ICT facilities for HPC that are scalable
and flexible as well as compliant with regulations and guidelines regarding privacy protection and
FAIR data management. An HPC program plan describes three principal solutions for HPC: the on‐
premises Research Zone, the Amsterdam UMC Research Cloud (through Surf Cumulus) and the Surf
HPC services. Realization of this detailed plan should ensure that from 2021 on HPC will be a basic
service for each researcher at Amsterdam UMC.
Acknowledgements not relevant
Keywords: HPC, Researchsupport
Abstractbook Health‐RI conference 2020
Abstract: 58 (Poster)
EPTRI ‐ European Paediatric Translational Research Infrastructure: a
bridge towards the future of paediatric medicine Tessa van der Geest (1), Valery Elie (2), Miriam G. Mooij (1, 3), Donato Bonifazi (4), Doriana
Filannino (4), Annalisa Landi (5), Mariangela Lupo (6), Lucia Ruggieri (5), Ales Stuchlik (7), Evelyne
Jacqz‐Aigrain (2), Saskia N. de Wildt (1, 8)
(1) Department of Pharmacology and Toxicology, Radboud University Medical Center, Nijmegen, The
Netherlands, (2) Paris Diderot University ‐ University Hospital Robert Debré – Paediatric Pharmacology and
Pharmacogenetics, 48 boulevard Sérurier ‐ 75019 Paris, France, (3) Department of Pediatrics, Leiden University
Medical Center, Leiden, The Netherlands, (4) Consorzio per Valutazioni Biologiche e Farmacologiche, Via
Putignani 178 ‐ 700122 Bari, Italy, (5) Gianni Benzi Pharmacological Research Foundation, Via Putignani, 133 ‐
70121 Bari, Italy, (6) TEDDY European Network of Excellence for Paediatric Clinical Research, Via Luigi Porta 14 ‐
27100 Pavia, Italy, (7) Institute of Physiology, Czech Academy of Sciences, Prague, Vídeňská 1083 ‐ 142 20
Praha, Czech Republic, (8) Intensive Care and Department of Paediatric Surgery, Erasmus MC Sophia Children’s
Hospital, Rotterdam, the Netherlands.
The European Paediatric Translational Research Infrastructure (EPTRI) aims to propose
developmental models for a future basic Research Infrastructure (RI) fostering high level basic and
applied research from drug discovery to paediatric formulation. The future RI will be complementary
to the existing RIs by putting together and networking all the available competences and
technologies useful to improve paediatric research in paediatric medicines.
For this purpose, EPTRI is preparing a Conceptual Design Report (CDR) describing the scientific and
technical requirements as well as the key components of the future RI. This CDR will represent a
relevant strategy for the design and future set‐up of the new RI. In addition, the project covers the
main areas of need in paediatric medicines technology, creating five technical and scientific domains
including 4 thematic platforms: 1) paediatric medicines discovery; 2) paediatric biomarkers and
biosamples; 3) developmental pharmacology; 4) paediatric medicines formulations and medical
devices (see Figure 1 and 2); and the scientific domain underpinning medicines development to
paediatric studies. EPTRI is coordinated by Consorzio per Valutazione Biologiche e Farmacologiche
(CVBF) and involves 29 partners from 21 EU/Associated countries including existing RIs and the major
paediatric expertise to cover the scientific topics in the proposal. Moreover, EPTRI has received
relevant support from several national/regional authorities, patients associations, academy and
health institutions, thus demonstrating a favourable framework for the future technology‐driven RI
focused on paediatrics.
Creating a framework for a future paediatric RI will help to accelerate paediatric drug development
processes, resulting in a substantial improvement of children’s quality of life. EPTRI will allow to
increase knowledge and research within the technical and scientific domains and facilitate transfer of
innovations to the clinics for the benefit of children.
Acknowledgements This project has received funding from the European Union’s Horizon 2020 Research and Innovation
Programme under Grant Agreement n. 777554.
Keywords: Key Words: Paediatric medicines; research infrastructure; children; biomarkers;
pharmacology; formulation
Abstractbook Health‐RI conference 2020
Abstract: 59 (Poster)
Public‐private partnerships in biobanking and biobank‐related
research Van der Stijl, R. (1, 4), Manders, P. (2), Scheerder, B. (1), Broeks, A. (3), Schaaij‐Visser, T. ( 4), van
Nuland, R. (5), Eijdems, E.W.H.M (1, 4)
(1) University Medical Center Groningen, (2) Radboud University Medical Center, (3) Netherlands
Cancer Institute, (4) BBMRI‐NL, (5) Health‐RI
Introduction
Biobanks and similar research infrastructures are responsible for their own sustainability. However,
they are also dependent on their surrounding macro‐environment. We have to create an
environment that enables and promotes sustainable biobanking. To do this we should strive for
appropriate overarching preconditions on a legislative, policy, organisational, and financial level. As a
first step we gathered input from biobanks and their users on current challenges and possible
solutions. This input from the supply and demand perspective is a starting point for further
discussions with relevant stakeholders.
Methods
We organised a workshop with 20 biobanks and data infrastructures. In addition, we organised four
focus groups with biobank users from academia and pharmaceutical industry.
Results
Biobanks and users indicated facing challenges on quality; accessibility; visibility; ethical, legal, and
societal issues; and financing. Sample and data issuance policies are complex and differ across
institutions. In addition, it became clear that the current incentives do not promote sharing and
sustainability. Further improvements could be made on the image of biobanks and clarity about their
impact. The potential for public‐private partnerships in biobanking and biobank‐related research
could be better utilized by bringing both parties to the table at an earlier stage. All parties indicated
the growing importance of data for research.
Discussion
The input of both biobanks and biobank users will be combined into recommendations for
overarching preconditions aimed to increase the use and sustainability of biobanks and similar
infrastructures. Setting suitable preconditions is only possible through the combined actions of all
stakeholders, including biobanks, researchers, research institutions, policymakers, and funders. Only
by collaborating can we work towards sustainable solutions, for the benefit of medical research,
health care, and the Dutch population.
Acknowledgements We would like to acknowledge the input of all workshop and focus group participants
Keywords: Sustainable biobanking, biobanks, stakeholders, recommendations
Abstractbook Health‐RI conference 2020
Abstract: 60 (Poster)
Recommendations for sustainable biobanking Van der Stijl, R. (1, 2), Scheerder, B. (1), Eijdems, E.W.H.M (1, 2)
(1). University Medical Center Groningen, (2) BBMRI‐NL
Introduction
To be sustainable biobank and other similar research infrastructures need to be operative, effective,
and competitive over their expected lifetime. However, sustainability is complicated and requires
finding the right balance on a social, financial, and operational level; within a dynamic environment.
Sample and data infrastructures can improve their sustainability by following these nine
recommendations, in the context of their own individual situation.
Methods
Through literature research, case study analysis, workshops, and focus groups we were able to
extract good practices and translate these into nine recommendations, with a focus on the financial
dimension, which might help biobanks in their search for sustainability.
Results
The nine recommendations are:
1. The right start by drafting a business plan
2. Adopt a user‐centred perspective
3. Know and show your value
4. Choose a business model and stick to it
5. Get a grip on your costs
6. Find multiple sources of funding
7. Engage with your key stakeholders
8. Become an attractive partner to industry
9. Make sure samples and data are (re)used
Discussion
There are things to learn from the business world. However, academic research infrastructures
should not forget that they are primarily a not‐for‐profit endeavour, operating in a complex medical‐
ethical environment. As there is a large diversity in biobanks and similar research infrastructures
there are no one‐size‐fits‐all solutions. These recommendations should therefore be applied from
one’s own perspective.
Acknowledgements We would like to acknowledge the support from Health‐RI, BBMRI‐NL and our BBMRI‐NL WP6
members. In particular, we want to acknowledge Tieneke‐Schaaij Visser for providing input and
support, and Rick van Nuland for supporting our workshops and focus groups.
Keywords: Sustainability, research infrastructure, recommendations, sustainable biobanking,
biobank
Abstractbook Health‐RI conference 2020
Abstract: 61 (Poster)
COREON – Committee on Regulation and Research P. Manders (1), M.K. Schmidt (2), M. Paardekooper (3), E.J. de Gaag (4), E.B. van Veen (5), L.M.
Bouter (6)
(1) Radboud Biobank, Radboudumc, Nijmegen, The Netherlands, (2) Division of Molecular Pathology,
NKI‐AVL, Amsterdam, The Netherlands, (3) EMGO+ Institute, Amsterdam UMC, Amsterdam, The
Netherlands, (4) Pharmo Institute, Utrecht, The Netherlands, (5) MLC Foundation, Den Haag, The
Netherlands, (6) Department of Epidemiology and Biostatistics, Amsterdam UMC, Amsterdam, The
Netherlands
Why?
The goal of COREON, the Committee on Regulation and Research of the Federa, is to encourage
careful and responsible research with health data and human tissues, aiming for a balance between
the public interest in such research, the participants involved and those of researchers. COREON
positions itself as the intersection between observational researchers and the regulation of such
research and as the platform where such issues are discussed amongst researchers.
Who?
COREON was established 2003, first as subcommittee of the VvE (the Dutch Epidemiological Society),
and later under the umbrella of the Federa (www.federa.org). COREON consists of people who are
active within observational health research. They represent a broad range of Dutch academic and
other research centres. The activities of COREON are funded by annual contributions from the
participating organizations.
What and how?
Via its plenary meetings (3‐4 times a year) COREON participants discuss scientific, legal and ethical
issues of observational research. Each year COREON organises a session on WEON, the yearly
convention of epidemiological researchers. COREON initiates working groups on specific questions
and publishes statements to guide researchers. COREON was responsible for the Code of Conduct on
health research with patient data of 2004 and the Code of Conduct on responsible use of human
tissue of 2011. COREON commissioned the revision of both Codes of Conduct and will be responsible
to submit the new Code of Conduct on health research with patient data for approval to the Dutch
Data Protection Authority.
Acknowledgements ‐
Keywords: Regulation, observational research, code of conduct, health data, human tissue
Abstractbook Health‐RI conference 2020
Abstract: 62 (Poster)
Galaxy in education using the SURF Research Cloud M. J. Brandt (1), J. Koehorst (2), I. Nooren (1)
SURF (1), Wageningen University and Research (2)
At Wageningen University and Research (WUR) the Galaxy environment is used as a tool for
education to give students gives access to data intensive biomedical research tools in a user‐friendly
environment. For the environment it would be an advantage to be accessible from any computer by
starting it in the cloud. Setting up a Galaxy cloud environment for a group of students can
complicated and repetitive work.
The SURF Research Cloud (RSC) is a SURF service to make using the cloud easier for the SURF
members, ea. the Dutch research and education institutes. What RSC adds as a value to the existing
research cloud offerings is that it presents a single‐entry point to all cloud needs of a researcher with
all the state‐of‐the‐art technologies tied together. In addition to in‐house SURF IAAS offering “HPC‐
Cloud” and connections to other institute clouds, many of the available public cloud platforms such
as AWS and Azure are integrated to RSC allowing the existing users of these platforms manage their
applications with just a few clicks through the RSC portal.
For a research application like the Galaxy environment the RSC platform offers building blocks such
as starting with specific configuration and tool set every time, inviting users using their federated
identity and linking to datasets and persistent storage. This means the Galaxy service only has to be
configured once and can be restarted through our user‐friendly portal with a few clicks. Managing a
separate server is not needed anymore.
The users of the Galaxy environment can log in to the portal with institute account and start using
fully setup Galaxy environment by pressing the access button.
The next step after this pilot is to connect a fully configured Galaxy environment to scalable compute
for running more complex pipelines for research needs.
Acknowledgements SURF, Wageningen University and Research
Keywords: Galaxy, Cloud, data intensive biomedical research tools
Abstractbook Health‐RI conference 2020
Abstract: 63 (Poster)
European Paediatric Translational Research Infrastructure (EPTRI): a
survey to map the expertise of the excellence of developmental
pharmacology in pan‐European countries Tessa Van der Geest (1), Valery Elie (2), Miriam G. Mooij (1, 3), Donato Bonifazi (4), Doriana
Filannino (4), Annalisa Landi (5), Mariangela Lupo (6), Lucia Ruggieri (5), Ales Stuchlik (7), Evelyne
Jacqz‐Aigrain (2), Saskia N. de Wildt (1)
(1) Department of Pharmacology and Toxicology, Radboud University Medical Center, Nijmegen, The
Netherlands, (2) Paris Diderot University ‐ University Hospital Robert Debré – Paediatric Pharmacology and
Pharmacogenetics, 48 boulevard Sérurier ‐ 75019 Paris, France, (3) Department of Pediatrics, Leiden University
Medical Center, Leiden, The Netherlands, (4) Consorzio per Valutazioni Biologiche e Farmacologiche, Via
Putignani 178 ‐ 700122 Bari, Italy, (5) Gianni Benzi Pharmacological Research Foundation, Via Putignani, 133 ‐
70121 Bari, Italy, (6) TEDDY European Network of Excellence for Paediatric Clinical Research, Via Luigi Porta 14 ‐
27100 Pavia, Italy, (7) Institute of Physiology, Czech Academy of Sciences, Prague, Vídeňská 1083 ‐ 142 20
Praha, Czech Republic, (8) Intensive Care and Department of Paediatric Surgery, Erasmus MC Sophia Children’s
Hospital, Rotterdam, the Netherlands
INTRODUCTION: Currently, the European landscape related to the developmental pharmacology
appears scattered and with low awareness of available services and facilities in this field, resulting in
overlapping initiatives and inefficient use of financial, instrumental, and human resources. European
Paediatric Translational Research Infrastructure (EPTRI) aims to design the framework of a paediatric
Research Infrastructure (RI) intended to enhance technology‐driven paediatric drug discovery. Within
the project, 5 technical and scientific domains have been identified among which the developmental
pharmacology platform aimed to enhance knowledge on developmental changes affecting drug
disposition. We here present the developmental pharmacology platform
MATERIALS AND METHODS: Within EPTRI, a survey was launched among selected research centres in
the field of developmental pharmacology to map the expertise within paediatric pharmacology in
pan‐European countries and identify the possible gaps in the available paediatric research services
and facilities. Firstly, the survey was delivered to 74 recipients between April‐June 2018. Later on, to
have a wider map of the European research units and services, the survey was re‐opened and
distributed among 153 recipients between January‐April 2019.
RESULTS: 38 service providers answered the survey among which 8 came from UK, 7 from Italy, 6
from The Netherlands. The analysis allowed to define a map of services to be provided within the
developmental pharmacology platform and represented in Figure 1. Relevant expertise has been
identified such as analytical labs capable to set‐up sensitive drug assays, paediatric omics facilities,
pharmacometrics expertise, large databases adapted to paediatric pharmacoepidemiology, as well as
placental platforms.
CONCLUSION: This analysis allowed to map the research units and services that will be provided in
the field of developmental pharmacology platform within EPTRI. Likewise, it provided a point of
reflection for the scientific community on the strengths and weaknesses of this research areas and
the relevance of EPTRI to fill these gaps.
Abstractbook Health‐RI conference 2020
Acknowledgements This project has received funding from the European Union’s Horizon 2020 Research and Innovation
Programme under Grant Agreement n. 777554.
Keywords: Paediatric medicines; research infrastructure; children; biomarkers; pharmacology;
formulation
Abstractbook Health‐RI conference 2020
Abstract: 64 (Poster)
Thyroid function and metabolomics: from observational research in
BBMRI cohorts to causal inference through Mendelian Randomization Nicolien A. van Vliet (1), Maxime M. Bos (1, 2), Fariba Ahmadizar (2), Marian Beekman (3), Mariska
Bot (4), Layal Chaker (2, 5, 6), Christian Delles (7), Mohsen Ghanbari (2), Antonius E. van
Herwaarden (8), Evelyn Houtman (3), M. Arfan Ikram (2), Martin Jaeger (9, 10), J. Wouter Jukema
(11), Margreet Kloppenburg (12, 13), Jennifer Meessen (3, 14), Ingrid Meulenbelt (3), Yuri
Milaneschi (4), Simon P. Mooijaart (1), Mihai Netea (10), Romana Netea‐Maier (9), Robin P.
Peeters (5,6), Brenda Penninx (4), Naveed Sattar (7), Eline Slagboom (3), Carisha S. Thesing (4),
Stella Trompet (1), Raymond Noordam (1), Diana van Heemst (1), BBMRI Metabolomics
Consortium
(1) Section of Gerontology and Geriatrics, Department of Internal medicine, Leiden University Medical
Center, Leiden, the Netherlands, (2) Department of Epidemiology, Erasmus University Medical Center,
Rotterdam, The Netherlands, (3) Department of Biomedical Data Sciences, Section Molecular
Epidemiology. Leiden University Medical Center, Leiden, The Netherlands, (4) Department of
Psychiatry, Amsterdam Public Health research institute, Amsterdam UMC, Vrije Universiteit
Amsterdam, Amsterdam, the Netherlands, (5) Department of Internal Medicine and Academic Center
for Thyroid Diseases, Erasmus Medical Center, Rotterdam, the Netherlands, (6) Academic Center for
Thyroid Diseases, Erasmus Medical Center, Rotterdam, the Netherlands, (7) Institute of
Cardiovascular and Medical Sciences, College of Medical, Veterinary and Life Sciences, University of
Glasgow, United Kingdom, (8) Department of Laboratory Medicine, Radboud Laboratory for
Diagnostics (RLD), Radboud University Medical Center, Nijmegen, the Netherlands, (9) Department of
Internal Medicine, Division of Endocrinology, Radboud University Medical Center, Nijmegen, The
Netherlands, (10) Department of Internal Medicine and Radboud Center for Infectious Diseases,
Radboud University Medical Center, Nijmegen, The Netherlands, (11) Department of Cardiology,
Leiden University Medical Center, Leiden, the Netherlands, (12) Department of Rheumatology, Leiden
University Medical Center, Leiden, The Netherlands, (13) Department of Clinical Epidemiology, Leiden
University Medical Centre, Leiden, The Netherlands, (14) Department of Orthopaedics, Leiden
University Medical Center, Leiden, The Netherlands.
Thyroid hormones affect lipid metabolism, though it is unknown which specific lipid subclasses are
affected. Here we conducted an observational multi‐cohort study within the BBMRI framework and a
two‐sample Mendelian randomization (MR) study of thyrotropin (TSH) and free thyroxine (fT4) levels
within the reference range on Nightingale metabolomics.
We conducted an observational study in 6 cohorts (N=9,353) and MR analyses using published
summary‐level data from a genome‐wide association study (N=24,925). For the observational
analyses, we used linear regression adjusted for age, sex, BMI and smoking, subsequently meta‐
analyzed using random effects models. As genetic instruments for the MR studies we used 57 genetic
variants for TSH and 30 genetic variants for fT4 (explained variance 9.4% and 4.8% respectively).
Associations between the genetic instruments for TSH and for fT4 and the metabolites were modeled
using Inverse Variance Weighted (IVW). All analyses took into account multiple testing for 37
uncorrelated metabolites (P<1.34x10‐3).
Observationally, TSH was associated with 52/161 metabolite concentrations (mainly VLDL, fatty acids
and kidney function), and fT4 was associated with 21/161 metabolite concentrations across all
Abstractbook Health‐RI conference 2020
lipoprotein subclasses, fatty acids and ketone bodies. In the subset of 123 metabolites reported in
the data used for MR, genetically determined higher TSH levels were associated with lower
concentration of very large HDL only (IVW ‐0.09 SD, 95% C.I. ‐0.14;‐0.05, P=1.66x10‐4), while
genetically determined higher fT4 levels were associated with higher glycoprotein acetyls only (IVW
0.12 SD, 95% C.I. 0.05;0.19, P=7.88x10‐4). Sensitivity analyses yielded similar results.
Variation in thyroid status within the reference range is associated with a distinct metabolic profile,
though causality is not yet ascertained. Possible explanations for the discrepancy in the results
between the observational and MR analyses include differences in power, residual confounding or
different biological mechanisms for the observed compared to the genetically determined thyroid
status.
Acknowledgements This work was performed within the framework of the BBMRI Metabolomics Consortium funded by
BBMRI‐NL, a research infrastructure financed by the Dutch government through Netherlands
Organisation for Scientific Research (NWO) (Grant Nos. 184.021.007 and 184033111) and supported
by the European Commission project THYRAGE (Horizon 2020 research and innovation programme,
666869).
Keywords: BBMRI, thyroid, metabolomics, lipoproteins, Mendelian randomization
Abstractbook Health‐RI conference 2020
Abstract: 65 (Poster)
Applying the FAIR Data principles to a Rare Disease registry: a case
study of the VASCA registry Bruna dos Santos Vieira (1, 2), Karlijn Groenen (1), Annika Jacobsen (3), Martijn G. Kersloot (4, 5),
Rajaram Kaliyaperumal (3), Ronald Cornet (4), Peter A. C. ’t Hoen (2), Marco Roos (3), Leo Schultze
Kool (1, 6)
(1) Dept. of Radiology and Nuclear Medicine, Radboud university medical center, Nijmegen, The
Netherlands, (2) Center for Molecular and Biomolecular Informatics, Radboud university medical
center, Nijmegen, The Netherlands, (3) Dept. of Human Genetics, Leiden University Medical Center,
Leiden, The Netherlands, (4) Amsterdam UMC, University of Amsterdam, Dept. of Medical
Informatics, Amsterdam Public Health Research Institute, Amsterdam, The Netherlands, (5) Castor
EDC, Amsterdam, The Netherlands, (6) VASCERN VASCA European Reference Centre
Registries of rare disease (RD) patients are extremely useful for establishing genotype‐phenotype
relationships, natural history studies and selection of patients for clinical trials. Each of the (local)
registries for a given disease may only contain a limited number of patients. Analyses across different
RD registries would help to increase patient numbers, but these analyses are usually difficult because
each registry is set up differently and data may not be accessed or at least not in a uniform way.
Therefore, applying FAIR data principles to RD registries is vital. Aiming at increasing FAIRness among
registries, the Platform on Rare Disease Registration (EU RD Platform) defined a set of Common Data
Elements (CDEs). This abstract describes how we implemented the CDEs and the FAIR data principles
in the Registry of Vascular Anomalies (VASCA).
A semantic model defining the CDEs and the relationships between them was created, adapted and
finalized by peer feedback. This model was then transformed into a Resource Description Framework
(RDF) template. Subsequently, an application was developed to feed a Twig template, which in turn
populates the RDF template with data entered in Castor EDC’s electronic Case Report Form (eCRF).
The RDF is accessible within a FAIR Data Point (FDP), allowing researchers to query and re‐use the
data in real‐time. During this implementation, several stakeholders were involved including patient
organization, domain‐, data‐ and ontology experts. Currently, VASCA is published in the EU RD
Platform metadata repository (ERDRI.mdr) and directory of registries (ERDRI.dor).
In conclusion, we successfully set up the infrastructure for a FAIR RD registry based on the CDEs. The
next step entails actual data collection within the participating centers. Also, we will investigate
interoperability by performing federated SPARQL queries between multiple registries.
Acknowledgements One author of this abstract is a member of the Vascular Anomalies Working Group (VASCA WG) of
the European Reference Network for Rare Multisystemic Vascular Diseases (VASCERN) ‐ Project ID:
769036.
Keywords: Vascular anomalies, rare disease registry, federated queries, RDF, SPARQL, Twig
template, FAIR data, common data elements, data model
Abstractbook Health‐RI conference 2020
Abstract: 66 (Poster)
The metabolic profile of arterial calcification in the multi‐cohort
BBMRI setting Maxime M. Bos (1), Nicolien A. van Vliet (2), Marian Beekman (3), Eline Slagboom (3), BBMRI
Metabolomics Consortium, Meike Vernooij (4), Jeroen van der Grond (5), Fariba Amahdizar (1),
Mohsen Ghanbari (1), Arfan Ikram (1), Diana van Heemst (2), Daniel Bos (1), Maryam Kavousi (1)
(1) Department of Epidemiology, Erasmus University Medical Center, Rotterdam, The Netherlands,
(2)Section of Gerontology and Geriatrics, Department of Internal medicine, Leiden University Medical
Center, Leiden, The Netherlands, (3) Department of Biomedical Data Sciences, Section Molecular
Epidemiology, Leiden University Medical Center, Leiden, The Netherlands, (4) Department of
Radiology, Erasmus University Medical Center, Rotterdam, The Netherlands, (5) Department of
Radiology, Leiden University Medical Center, Leiden, The Netherlands
Increasing evidence shows that greater arterial calcification leads to elevated risk of atherosclerotic
cardiovascular disease. However, the underlying biological mechanism of site‐specific calcification is
largely unknown. Within the BBMRI framework, we performed a multi‐cohort study on the
associations of the metabolic profile with calcification of coronary arteries (CAC), aortic arch (AAC)
and the aortic valve (AVC).
We included a total of 1114 participants from the population‐based Rotterdam Study and 390 from
the Leiden Longevity Study. Blood samples were used to determine a wide range of plasma
metabolites by proton nuclear magnetic resonance (NMR). Participants underwent non‐contrast
computed tomography to quantify CAC, AAC and AVC. Linear regression modelling adjusted for
relevant covariates was used to assess the associations of 166 metabolites with CAC, AAC and AVC.
Correction for multiple testing was based on 33 independent metabolites (p‐value 0.05/33 = 1.5 x 10‐
3).
One standard deviation (SD) increase in concentration of a1‐acid glycoprotein, was associated with a
0.10 SD increase in AAC (standard error (SE) = 0.03, p‐value = 9.5 x 10‐4). When considering sex‐
specific effects, we observed an association of acetate with CAC (beta = ‐0.09, SE = 0.03, p‐value = 4.1
x 10‐4) in women.
Higher levels of circulating glycoproteins were associated with increased AAC. Moreover, acetate was
associated with CAC only among women indicating differences in metabolic profile of CAC between
men and women. These results provide evidence for location‐specific differences and sex‐specific
effects in etiology of atherosclerosis.
Acknowledgements This work was performed within the framework of the BBMRI Metabolomics Consortium funded by
BBMRI‐NL, a research infrastructure financed by the Dutch government through Netherlands
Organisation for Scientific Research (NWO) (Grant Nos. 184.021.007 and 184033111). NvV and DvH
were supported by the European Commission project THYRAGE (Horizon 2020 research and
innovation programme, 666869).
Keywords: BBMRI, arterial calcifications, metabolomics, multi‐cohort, cardiovascular diseases
Abstractbook Health‐RI conference 2020
Abstract: 67 (Poster)
e/MTIC ‐ Health Data Portal initiative (1) Eindhoven University of Technology, (2) Catharina Hospital, (3) Kempenhaeghe Epilepsy and Sleep
Center (4) Royal Philips Eindhoven, (5) the Maxima Medical Center
Abstract e/MTIC ‐ Health Data Portal initiative
The Eindhoven MedTech Innovation Center (e/MTIC) is a large‐scale research collaboration aimed at
improving public healthcare through high‐tech health innovations.
Within this consortium with Catharina Hospital, the Maxima Medical Center, Kempenhaeghe Epilepsy
and Sleep Center, Eindhoven University of Technology and Royal Philips Eindhoven, we are
developing the Health Data Portal to facilitate and enable joint research projects.
The Health Data Portal (HDP) is a scalable collaboration platform that builds on existing initiatives to
provide an infrastructure where researchers can bring together and work safely with medical data. It
brings together medical institutes, academia and commercial partners to provide a fast track to
innovation.
In the design of the HDP, we have given highest priority to GDPR‐compliancy, without compromising
the research requirements. By using a trusted organisation as independent third party to process and
handle data requests, we can build trust and confidence that records are held securely and data is
de‐identified appropriately. The HDP also provides secure solutions for medical studies that require
combining data‐sets from multiple institutions and sources, stimulating collaboration across.
Healthcare innovations are increasingly based and depending on data‐driven research.
The HDP will gradually build up a rich metadata catalogue of medical information from the e/MTIC
partners, first within the domains of cardiovascular‐care, perinatal‐care and sleep‐care. This central
catalogue can grow by including other health domains and institutes in the future.
These large medical datasets, available between the data science platform members, will enable
scientists to develop and test new hypotheses about human health. The data sets can enable new AI
applications for machine learning to revolutionize healthcare.
We have the ambition to grow into a national platform that can lead to faster life sciences and health
innovation.
Acknowledgements The Eindhoven MedTech Innovation Center (e/MTIC)
Keywords: IT Architecture, Healthcare, Portal, Platform, Collaboration, Data analysis, Data Science,
Artificial Intelligence (AI), Anonymization, Pseudonymization, Medical Engineering, Diagnostics,
Value‐based.
Abstractbook Health‐RI conference 2020
Abstract: 68 (Poster)
The Pregnancy And Childhood Epigenetics (PACE) Consortium ‐ A
platform for epigenome‐wide association meta‐analyses Janine F. Felix (1,2), Stephanie J. London (3), on behalf of the PACE Consortium
(1) The Generation R Study Group, Erasmus MC, University Medical Center Rotterdam, Rotterdam, the
Netherlands, (2) Department of Pediatrics, Erasmus MC, University Medical Center Rotterdam,
Rotterdam, the Netherlands, (3) Epidemiology Branch / Genetics, Environment & Respiratory Disease
Group, National Institute of Environmental Health Sciences, Research Triangle Park, NC, USA
Differential DNA methylation represents a potential mechanism underlying associations of early‐life
exposures and later‐life health. In recent years, many pregnancy, birth and childhood studies have
initiated research on DNA methylation, using Illumina 450K or 850K EPIC arrays. These data can be
used in epigenome‐wide association studies. As individual studies are usually underpowered for such
studies, collaboration between studies and combined meta‐analyses are needed to optimize the use
of resources and increase the likelihood of detecting DNA methylation differences.
The global PACE Consortium brings together 40 studies with DNA methylation data in over 30,000
pregnant women, newborns, and children. Its primary aim is to identify differential DNA methylation
related to exposures and outcomes pertinent to health in pregnancy and childhood through joint
analysis of DNA methylation data. Secondary aims are to perform functional annotation‐based
analyses, to study causality of DNA methylation differences for child health phenotypes, to
contribute to methodologic development, and to exchange knowledge and skills. Studies include
participants from various backgrounds in terms of ethnicity, age, and living environment, enabling
testing of identified associations across different settings.
Findings to date include associations of prenatal maternal smoking, body mass index and air
pollution exposure with offspring DNA methylation at birth and in childhood, as well as associations
of DNA methylation with child asthma and lung function. Ongoing work focuses on further
gestational exposures, such as maternal stress and nutritional exposures, as well as child health
outcomes including cardio‐metabolic and neuro‐developmental phenotypes. A current overview of
published papers can be found at:
http://www.niehs.nih.gov/research/atniehs/labs/epi/pi/genetics/pace/index.cfm. The PACE
Consortium is an open, dynamic collaboration. Additional research groups are welcome to join. It
offers a strong platform to study the role of DNA‐methylation in the associations of early‐life
exposures and later health outcomes and to contribute to the field of population epigenetics.
Acknowledgements NA
Keywords: DNA methylation, consortium, epigenetics
Abstractbook Health‐RI conference 2020
Abstract: 69 (Poster)
The DANS services for sharing, cataloguing and archiving your health
data Cees Hof, Ingrid Dillo, Heidi Berkhout
Data Archiving and Networked Services (DANS)
DANS (Data Archiving and Networked Services) is the Netherlands institute for permanent access to
digital research resources. DANS encourages researchers to make their research data and related
digital outputs Findable, Accessible, Interoperable and Reusable (FAIR). To realise our mission, DANS
provides expert advice and certified services. DataverseNL is the DANS service for short‐term data
management, EASY our long‐term data archive, and NARCIS the national catalogue service for
scholarly information. Training and consultancy services are provided for generic Research Data
Management and Data Management Planning. More specific training sessions focus on repository
certification, metadata standards, software sustainability and knowledge organisation systems. The
(coordinating) activities of DANS in (inter)national projects and networks, ensure constant innovation
and a state‐of‐the‐art knowledge on infrastructural data developments.
Although the roots of DANS are within the humanities and social sciences, most DANS services are
generic services relevant for nearly all scientific disciplines, including the life and health sciences. As
part of the Dutch national e‐infrastructure for research data, DANS is involved in several projects and
initiatives around health data, often acting at the cross roads between the life and social sciences.
Also, the DANS training activities touch upon the developments around health data. Cataloguing the
Dutch “zorggegevens” in NARCIS, or the DANS training modules in the Helis Academy FAIR data
stewardship course, are examples of specific DANS contributions to the life and health sciences.
The DANS poster presentation provides an overview of the DANS services of interest to the owners
and custodians of health data, including examples of relevant recent projects. DANS invites
participants of the Health‐RI 2020 conference to probe how DANS could support the sharing,
cataloguing and archiving of their health data.
Acknowledgements ‐
Keywords: FAIR data, services, training, archiving, data sharing, data catalogue
Abstractbook Health‐RI conference 2020
Abstract: 70 (Poster)
Metabolic risk scores: from metabolome to phenotype and back D. Bizzarri (1, 2), M.J.T. Reinders (2, 3), P.E. Slagboom (1,4), E.B van den Akker (1, 2, 3)
1) Molecular Epidemiology, LUMC, Leiden, The Netherlands, 2) LCBC, LUMC, Leiden, The Netherlands,
3) Delft Bioinformatics Lab, TU Delft Delft, The Netherlands, 4) Max Planck Institute for the Biology of
Ageing, Cologne, Germany
Introduction: The blood metabolome incorporates cues of the environmental and genetic
background of an individual, potentially offering a holistic view of its health status. Different types of
diseases have similar impacts on the blood metabolome, hence, that the blood metabolome might
not be disease specific. With this premises, we tried to identify novel metabolic states representing
the risk for multiple related cardio‐metabolic outcomes, using metabolic predictors for biomarkers
typically used in the clinic.
Methods: We will use the data available in BBMRI‐NL (composed by 1H‐NMR serum metabolomics
for 29 cohorts), to investigate the metabolic component of the available risk factors. We will use a
penalized regression model, to automatically select a subset of metabolites whose linear
combination will best predict each risk factors. We will employ two evaluation procedures: a 5‐Fold‐
Cross‐Validation and a Leave‐One‐Biobank‐Out‐Validation (holding out one cohort to use it as a test
set). From the trained models, we will obtain metabolic surrogate risk factors, which we will combine
training penalized Cox regression models, on 2 cohorts (enriched with cardiometabolic diseases) to
predict different cardio‐metabolic status of these individuals.
Results: In an exploratory analysis, we investigated which penalized regression method could deliver
the best metabolic prediction, using the data of one of the BBMRI cohorts (LLS‐P/O). We performed a
Two‐Deep‐Cross‐Validation analysis for 12 risk factors (both continuous and binary) to evaluate the
accuracy of: Ridge (RR), Lasso (LR) and ElasticNET (EN) regression. In these settings, we obtained
similar accuracy scores for the three methods, and we obtained good performances in particular for
the prediction of gender (auc~0.92), type2 diabetes (auc~0.88), statins use (auc~0.76).
Conclusions: We observed that the 1H‐NMR blood metabolomics could be used to accurately predict
several clinically relevant variables using penalized regression models.
Acknowledgements MOLEPI, LCBC, BBMRI‐NL
Keywords: metabolome, risk factors, metabotypes, BBMRI‐NL, regression models
Abstractbook Health‐RI conference 2020
Abstract: 71 (Poster)
Towards precision diagnostics: Untargeted metabolomics for the
diagnosis of inborn errors of metabolism in individual patients Purva Kulkarni 1, Albert Gerritsen 1,2, Udo F.H. Engelke 1, Brechtje Hoegen 2, Siebolt de Boer 1, Ed
van der Heeft 1, Marleen C.D.G. Huigen 1, Leo A.J. Kluijtmans 1, Karlien L.M. Coene 1
(1) Department of Laboratory Medicine, Translational Metabolic Laboratory (TML), Radboud
University Medical Center, Geert Groote Plein Zuid 10, 6525, GA, Nijmegen, The Netherlands.
(2) Department of Human Genetics, Radboud University Medical Center, Nijmegen, The Netherlands.
Introduction
Inborn Errors of Metabolism are inherited conditions caused by genetic defects in enzymes or their
cofactors, resulting in a specific metabolite fingerprint showing accumulation of substrate or lack of
end‐product in patient body fluids(1). Untargeted metabolomics offers a comprehensive readout of
metabolic status on an individual patient basis. This makes it a promising tool for diagnostic
screening and treatment monitoring of IEM patients, especially when clinical presentations are non‐
specific.
Technological and methodological innovation
We have previously established Next‐Generation Metabolic Screening(2) as a metabolomics‐based
diagnostic tool for individual IEM‐suspected patients. To fully exploit the clinical potential of NGMS,
we have developed an automated computational pipeline to streamline analysis of complex data and
make it reproducible. The pipeline features a GUI that converts raw data, detects and aligns features
across samples and annotates them to identify significant deviations in patients as compared to
controls.
Results and impact
Using our automated computational pipeline, we have advanced the application of metabolomics in
clinical diagnostic setting to a next level. Our pipeline ensures reproducible and time‐efficient
metabolomics data management, processing and analysis. To validate this pipeline, we tested
samples of IEM patients, including several diagnoses that were not yet measured with NGMS, for
example L‐2‐hydroxyglutaric aciduria. Our results further expand the clinical applicability and IEM
portfolio of NGMS.
References
1. A. Tebani et al., International Journal of Molecular Sciences. 17 (2016)
2. K. L. M. Coene et al., Journal of Inherited Metabolic Disease. 41, 337–353 (2018)
Acknowledgements N/A
Keywords: metabolomics, diagnostics, rare diseases, big data, Inborn Errors of Metabolism, data
analysis, computational pipeline, Biomarkers, Mass spectrometry