Crossing the Chasm of Semantic Despair: Integrating ... the Chasm of Semantic Despair: Integrating Knowledge and Data from Science and Clinical Practice CG Chute, Johns Hopkins Clinical
47
Crossing the Chasm of Semantic Despair: Integrating Knowledge and Data from Science and Clinical Practice CG Chute, Johns Hopkins Clinical Research Forum – IT Roundtable Washington, DC 2 Nov 2017
Integrating Knowledge and Data from Science and Clinical Practice
CG Chute, Johns Hopkins
Clinical Research Forum – IT Roundtable Washington, DC
2 Nov 2017
On the Nature of Translational Research
“Translational science is the field of investigation focused on understanding the scientific and operational principles underlying each step of the translational process.” - NCATS web page
•Dimensions of Translation –Discovery –Dissemination –Acceptance –Adoption –Implementation
The 17th Anniversary of the “17 Years” Paper
Balas EA, Boren SA. "Managing clinical knowledge for health care improvement." Yearbook of medical informatics 2000 (2000): 65-70
Research translation takes 17 years, the message takes longer…
0
1
2
3
4
5
6
7
8
9
10
Biology Medicine
The Continuum Of Biomedical Informatics Bioinformatics meets Medical Informatics
5
The Chasm of Semantic Despair
Origins of Big Science: Astronomy
6
Sloan Digital Sky Survey III – DR9
Total area of imaging
31,637 square degrees
Image field size 1361x2048 pixels
Number fields 938,046 (excluding supernovae runs)
Catalog objects 1,231,051,050 Number of unique, primary sources Total 469,053,874 Stars 260,562,744 Galaxies 208,478,448 Unknown 12,682 7
Guoqian Jiang Clinical data standards & Interoperability, semantic
modeling & validation
Presenter
Presentation Notes
What strength(s) would you bring to this effort? The team is diverse in its data science expertise and spans basic research, clinical informatics, and computer science.
Building a Translational Bridge
CD2H
Presenter
Presentation Notes
We’ve been thinking about the need for combined mechanistic/molecular classification with phenomenological nosologies for some time. This figure is adapted from National Research Council (U.S.). Committee on A Framework for Developing a New Taxonomy of Disease., Toward precision medicine : building a knowledge network for biomedical research and a new taxonomy of disease. 2011, Washington, D.C.: National Academies Press. xiii, 128 p. http://www.nap.edu/catalog/13284/toward-precision-medicine-building-a-knowledge-network-for-biomedical-research Figure 3.1 (page 52): Building a biomedical Knowledge Network for basic discovery and Medicine.
1
2
3
Presenter
Presentation Notes
The proposed TransMed Data Translator is not dissimilar from the original NRC conception. This figure is adapted from National Research Council (U.S.). Committee on A Framework for Developing a New Taxonomy of Disease., Toward precision medicine : building a knowledge network for biomedical research and a new taxonomy of disease. 2011, Washington, D.C.: National Academies Press. xiii, 128 p. http://www.nap.edu/catalog/13284/toward-precision-medicine-building-a-knowledge-network-for-biomedical-research Figure 3.1 (page 52): Building a biomedical Knowledge Network for basic discovery and Medicine.
Leadership in existing projects to be leveraged in the translator
Deep open source experience, each step of the way
Dissemination / deployment
Extraction of knowledge from diverse sources
Data & knowledge integration,
algorithms & tools
Contribution in other projects
1
2
3
Leadership in other projects
OBO Foundry
Presenter
Presentation Notes
"Has your team contributed to open source projects in which they were not the project owner? Which ones?” This is not an exhaustive list; all participants contribute to numerous open source repositories in Github and Bitbucket. Bel – Hongfang Liu cTAKES – Hongfang Liu, Chris Chute Wikidata – Andrew Su, Ben Good Phenomizer - Peter Robinson Monarch Initiative – Melissa Haendel, Chris Mungall, Peter Robinson BioPortal – Chris Chute, Melissa Haendel, Chris Mungall, Hongfang Liu OBO Foundry – Melissa Haendel, Chris Mungall Global Alliance for Genomics and Health – Melissa Haendel, Chris Mungall, Peter Robinson, Shannon McWeeney ICD-11 – Chris Chute Exomiser – Peter Robinson, Melissa Haendel, Chris Mungall EVS - Chris Chute SHARPn – Chris Chute, Hongfang Liu Bioconductor - LexEVS – Chris Chute eagle-i – Melissa Haendel GMOD – Chris Mungall VIVO – Melissa Haendel Gatk – Shannon McWeeney BioPerl – Chris Mungall DeepDive – Andrew Su NCI QIN – Shannon McWeeney My* - Andrew Su, Ben Good OpenWetware – Maureen Hoatlin
My*
Tools to integrate vocabularies Tools to identify equivalencies: Identifier and synonym alignment Conceptual alignment based upon logic
determinations and prior probabilities Text mining for clinical concepts and
pathway fragments
(1) Extraction of knowledge from diverse sources TRANSMED KNOWLEDGE GRAPH
kBOOM
Presenter
Presentation Notes
What strength(s) would you bring to this effort? Our team has sophisticated expertise in addressing a wide range of data integration challenges and types.
(2) Data & knowledge integration, algorithms & tools
GRAPH INFERENCE OWL-based reasoning over large graphs BELpathway ‘chains of causation’ Probabilistic inference across disease-phenotype-
gene Bayesian Ontology Query Algorithm (Boqa)
GRAPH QUERY Query related entities within/across species,
sources Query similar sets of entities (OWLsim) Sets of phenotypes Expression patterns Pathway modules
Purpose built for clinicians, researchers, & bioinformaticians
TransMed outputs based on condition set and competency questions are iteratively compared against current phenomonological nosology
Comparisons of integrated delivery of mechanistic modules against single source inquiries
T-Score measuring connectivity of clinical and basic sources
Presenter
Presentation Notes
How will you validate the components and assess the performance of each demonstration project? For Fanconi Anemia and for a set of undiagnosed patients, we will iteratively compare the outputs of the TransMed platform to the current nosologies available for each condition. Given the rarity of the undiagnosed patients, we expect to require mechanistic insight to aid diagnosis and treatment selection. The question in these cases is, does the data in the TransMed platform provide additional insight beyond any of the individual sources? Meh edit: note that for FA there are already data for individuals with FA defects combined with aldehyde processing defects. These cases could serve as positive controls and tests for some of our analyses. E.g., with input of FA gene sequences, patient medical records, and phenotypic outcomes, would Translator find that the common aldehyde processing defect was a likely contributor to a more severe FA defect? What is missing from an analysis in the broader population?
Monarch (via Dipper) ingest so far in TransMed
http://bit.ly/monarch-data-dashboard
Presenter
Presentation Notes
Julie to make sankey with new sources? http://bit.ly/monarch-data-dashboard GINAS - chemical substances/drugs FAIRS-AEOLUS - drug side effects STRING - protein interactions JASPAR - transcription factor binding sites Reactome - gene-pathway associations Bgee - gene expression Uniprot-Ensembl integrations/enhancements ClinVar - variant - disease associations
Common mechanistic underpinnings of rare & common/complex disease
Neutral
Acetaldehyde metabolism mutations:
500 Million People Affected
Fanconi Anemia: 1 in 160,000 individuals worldwide
ALDH KO: ?
Rare, Catastrophic
Common, Subtle
Rare, Beneficial
Freq
uenc
y
Detrimental Beneficial
??
Presenter
Presentation Notes
Suggests that there are subtler variants with subtler phenotypes in the general population; knowing what these are can help inform people’s behavioral choices, both in pregnancy (fetal alcohol syndrome), but also in life (cancer susceptibility). Also suggests potential development of drug agonists for treatment. MEH edit: Since it is unknown how many individual carry the *combination* of FA and ADH genes (and environmental exposures) that result in a negative phenotype, this demonstrator is powerful because it aims to evaluate the combination effect of two genes—thus providing a platform to understand other common diseases that result from a combination of rare variants. How will your concept and your data sources help reveal problems with the current classifications? The Fanconi Anemia and Undiagnosed disease demonstrator will highlight the need for combining traditional phenomenological with mechanistic nosologies by inferring new classifications based upon subcellular, cellular, organ, and organismal functional attributes. Observations such as the fact that Asian Flushing disease increases likelihood of squamous cell carcinoma can lead to hypotheses and candidates for drug repurposing.
Fanconi mechanistic inference
Presenter
Presentation Notes
How will your concept and your data sources help reveal problems with the current classifications? The Fanconi Anemia and Undiagnosed disease demonstrator will highlight the need for combining traditional phenomological with mechanistic nosologies by inferring new classifications based upon subcellular, cellular, organ, and organismal functional attributes. As shown in blue on this slide: - Reactive aldehydes are generated during normal metabolic processes, so organisms have developed a series of aldehyde dehydrogenases (ALDH) for detoxification. - DNA damage caused by excess aldehyde buildup is repaired by the proteins in the Fanconi anemia pathway. - If the FA pathway is defective, a specific type of DNA repair is defective. As shown in green: - New research in mice has shown that if both FA and ALDH are defective, then the phenotype is much more severe. Approx 500 million to 1 billion individuals have a variant in ALDH2 that causes a defect in detoxifying alcohol (Asian Flushing Syndrome). Cases of idiopathic bone marrow failure have been linked to this variant, and alcohol use in individuals with this variant have increased risk of squamous cell carcinoma. These studies support the interaction of the ALDH and FA pathway outcomes observed in mice.
The Fanconi Anemia mechanism
Fanconi Phenotypes
Presenter
Presentation Notes
Med edit: This slide underscores the idea that the FA phenotypes can be viewed as an exaggerated early aging process.
Each person - one sentence description, primary interest in ref: to CD2H John Wilbanks is the Chief Commons Officer at Sage Bionetworks, and his work there on data governance, engagement, and outreach will inform CD2H projects. Sean Mooney is the CRIO of UW Medicine and is heavily involved in a number of NIH networks and initiatives. Here he will provide high level oversight of the CD2H with other PIs and will be involved with IT platforms that support and enable CTSAs (software) Christopher Chute is is a Bloomberg Distinguished Professor at Johns Hopkins University, physician-scientist, and biomedical informatician and serves as the Chief Health Research Information Officer of Johns Hopkins Medicine. Here he leads clinical data and will guide efforts related to biomedical terminologies and health information technology standards. Kristi Holmes is the Director of Galter Health Sciences Library and directs evaluation for the Northwestern University Clinical and Translational Sciences Institute and several other large centers and programs. Here she will lead data analytics and continuous improvement efforts and contribute to other program efforts.
Presenter
Presentation Notes
These are the CD2H partners, but this is about you, about the CTSA programs, about the other things on your campus, other consortiums, international movements.
How do we achieve this?
CD2H
We collectively have a remarkable opportunity to build upon the excellence of the CTSA Program to innovate, pilot, implement, and sustain new informatics technologies to improve human health
Presenter
Presentation Notes
Also not just about informatics but anything that is enables informatics, such as data sharing policies and processes defines framework around which we are trying to build tools, etc. How are we going to do this. We are open to all We support open data, software and science, while enabling privacy and security We aspire to empower individual biomedical informatics programs to be agents of influence and change We support a champion model, where individual projects are proposed and led by champions We believe in transparency and shared governance We support all informatic areas of need within the CTSA network
Idea-to-Implementation Pipeline
CD2H
Presenter
Presentation Notes
Something here to show that this is a bipartite mission.. How will we achieve this? A combination of cultural and technological steps to empower the community to innovate, pilot, implement, and sustain new informatics the CTSA Program has a long history of innovative tools and methods and we are excited to build on this strong tradition with you. TECHNOLOGY: previous aims CULTURE: Amplify the community of practice - via collaborations (Rare disease and lifespan, I2I, DREAM Challenges), training, outreach, and evaluation OLD NOTES: We think of this as the Idea-to-Implementation collaborative pipeline While not necessarily created by the CTSAs, many tools have been influential through the CTSA network, including REDCap, i2b2, TranSMART, CTMS systems, etc. We will embrace new and innovative tools and methods
Each CTSA is a workbench
CD2H
1: Harmonize the data ecosystem
2: Realize a software tool ecosystem
3: Synthesize a people ecosystem
4: Catalyze technical and cultural evolution
Presenter
Presentation Notes
Goldilocks solutions to data dissonance; development of Good Data Practice (GDP); promotion of software standards for interoperability. [here maybe say boards are measured to nearest ⅛ inch, grouped by type of wood, not grouped according to whether the tree was felled on tuesday or wednesday] Apply standards to harmonize and integrate data. Improve navigation of the data sharing odyssey.
✓Goldilocks solutions to data dissonance; development of Good Data Practice (GDP); promotion of software standards for interoperability
✓Apply standards to harmonize and integrate data
✓Improve navigation of the data sharing odyssey
Aim 1: Harmonize the data ecosystem
CD2H
Presenter
Presentation Notes
image not licensed
✓Assemble communities of practice ✓Compile software best practices and
standards ✓Configure Synapse Platform for data
sharing and provide training ✓Deploy CIELO: a software
“storefront” ✓Verify & validate existing/emergent
software
Aim 2: Realize a software tool ecosystem
CD2H
Presenter
Presentation Notes
image at top is via pixabay, image at bottom is licensed from adobe stock Assemble I2I communities of practice for software development and assess their needs Create a compendium of best practices and standards for software development Configure Synapse Platform for I2I groups and provide training Deploy CIELO: a “storefront” for the discovery, adoption, and adaptation of software Verification and validation of existing and/or emergent software tools
✓Extend representation of expertise and services across the CTSA consortium
✓ Enable discovery of services/ data/tools/ people across the CTSA consortium
Aim 3: Synthesize a people ecosystem
CD2H
Demonstration disease areas: ✓Rare disease and eliminating the diagnostic odyssey ✓Improve the impact of clinical research across the lifespan
Aim 4: Catalyze technical & cultural evolution
CD2H
Presenter
Presentation Notes
Merge with “how will we achieve this
Clinicians/ care team Pathologists Ontologists Informaticians Curators Basic
Research
The translational workforce: It takes a village to solve disease
Informatics & data science competencies Mapping across current competency efforts (AMIA, CTSAs, ISCB, AMSTAT, etc)
Interdisciplinary collaboration, open science,
& entrepreneurship Partnerships with CTSA Workforce
Development, AMIA Informatics Educators and others
Mentoring network Synergize w NIH, BD2K Data Science RoAD-Trip, AMIA Mentor Match etc.
“Packaging” of Learning Modules into LMS Collaboration with N-lighten to get training modules into learning management
systems (LMS)
Module annotations Improve discovery/reuse
coordinate BD2K erudite, others
Presenter
Presentation Notes
Notes from Shannon: Should emphasize that this all aligns with NCATS Strategic Goal 3: Develop and foster innovative translational training and a highly skilled, creative and diverse translational science workforce (See objectives below) . Must be synergistic effort - I have highlighted some of key partners Objective 3-1: Identify and broadly disseminate the distinct knowledge, skillsets and core competencies needed by translational scientists. Objective 3-2: Develop and support translational science training and career development programs to cultivate a multi-faceted and diverse translational science workforce. Objective 3-3: Support the widespread acceptance of translational science as a formal discipline to help provide a clear and sustainable career path for translational scientists. Objective 3-4: Expand understanding of translational science through the deployment of cutting-edge communication and dissemination tools and technologies. notes from sean 1) Data science/informatics training material development 2) Technology for material deployment (LMS or MOOCs) 3) Contributing to the wrangling of workforce development community in the CTSA network with the aspirational goal of not creating 64 redcap training classes, for example. I've heard centralization, metadata and indexing, collaborative material development and a bit about platforms 4) Haven't heard mentor network other than in our proposal, but I personally like it.
Opportunities for CTSA Analytics & Data
CD2H
Optimize dissemination Dissemination and impact are inherently linked; build on best
practices for a range of outputs
Attribution Enable an attribution framework for
a range of products & processes for the translational workforce
Modular, dynamic evaluation library
Support local hubs in data-driven CQI through improved data, best
practices for analytics and data viz
Accountability Facilitate innovative accountability through open project dashboards
Collaborative Innovation Develop a model to support and grow CI across the CTSA Program
Presenter
Presentation Notes
A modular, dynamic evaluation library for hubs: best practices, dashboards, data visualization, resources for Common Metrics Program-wide merit-based peer review Enable an attribution framework for products & processes Empower broad dissemination Accountability to the CTSA Program through open dashboards Basically, I think something like this could work with images to communicate the following: a library of tools & resources (stack of books and papers?) peer review - maybe a magnifying glass? arrows pointed outward for dissemination something that looks like a dashboard center spot could be “CD2H” or something else Kristi to work on wording - mentioned social network analysis - “we are already doing this” Meets or exceeds best practice standards for quality control, accessibility, provenance, maintenance, support, and interoperability; Number of stakeholder-focused outputs; # of facilitated collaborative engagements; open licensing Develop Good Data Practice (GDP) FAIR-TLC maturity model for tools, ontologies, algorithms, data, software (TOADS); Success stories of implementing good data practices and software data standards; Impact on identified gaps in data ecosystem; Data harmony metrics metrics that measure collaboration, ascertainment and recruitment improved,
Continuing the Conversation
▪ Workgroups & collaborations ▪ Socio-technical communication environments ▪ Listservs, project email, and activities calendar ▪ We would love to come and visit your team! ▪ Interested in learning more? Please fill out the intake
