media.nature.com · web viewautomated database-guided expert-supervised orientation for...

Supplemental data

Automated database-guided expert-supervised orientation for immunophenotypic diagnosis and classification of acute leukemia

Ludovic Lhermitte1#, Ester Mejstrikova2#, Alita J. van der Sluijs-Gelling3,4#, Georgiana Emilia Grigore5,

Lukasz Sedek6, Anne E. Bras7, Giuseppe Gaipa8, Elaine Sobral da Costa9, Michaela Novakova2,

Edwin Sonneveld3, Chiara Buracchi7, Thiago de Sá Bacelar9, Jeroen G. te Marvelde7, Amélie

Trinquand1, Vahid Asnafi1, Tomasz Szczepanski10, Sergio Matarraz11, Antonio Lopez11, Belen

Vidriales12, Joanna Bulsa10, Ondrej Hrusak2, Tomas Kalina2, Quentin Lecrevisse11, Marta Martin

Ayuso5, Monika Brüggemann13, Javier Verde5, Paula Fernandez14, Leire Burgos15, Bruno Paiva15,

Carlos E. Pedreira16, Jacques J.M. van Dongen4, Alberto Orfao11$, and Vincent H.J. van der Velden7$,

on behalf of the EuroFlow Consortium

1Université Paris Descartes-Sorbonne Paris Cité, Institut Necker-Enfants-Malades, INSERM

UMR1151 and, Biological Hematology, AP-HP Necker-Enfants-Malades, 75015 Paris, France; 2CLIP -

Childhood Leukaemia Investigation Prague, Department of Paediatric Haematology and Oncology,

Second Faculty of Medicine, Charles University, University Hospital Motol, Prague, Czech Republic; 3Dutch Childhood Oncology Group, The Hague, Netherlands; 4Department of Immunohematology and

Blood Transfusion (IHB), Leiden University Medical Center, Leiden, The Netherlands; 5Cytognos SL,

Salamanca, Spain; 6Department of Microbiology and Immunology, Zabrze, Medical University of

Silesia (SUM), Katowice, Poland; 7Department of Immunology, Erasmus MC, University Medical

Center Rotterdam, Rotterdam, Netherlands; 8Tettamanti Research Center, Pediatric Clinic University

of Milano Bicocca,, Monza (MB),Italy; 9Department of Pediatrics, Faculty of Medicine, Federal

University of Rio de Janeiro, Rio de Janeiro, Brazil; 10Department of Pediatric Hematology and

Oncology, Zabrze, Medical University of Silesia (SUM), Katowice, Poland; 11Cancer Research Center

(IBMCC-CSIC), Department of Medicine and Cytometry Service, University of Salamanca (USAL) and

Institute of Biomedical Research of Salamanca (IBSAL), Salamanca, Spain; 12Hematology Service,

University Hospital of Salamanca, Salamanca, Spain 13Department of Hematology, University of

Schleswig-Holstein, Campus Kiel, Kiel, Germany; 14Institute for Laboratory Medicin, Kantonsspital

Aarau AG, Aarau, Swiss; 15Clinica Universidad de Navarra – Centro De Investigaciones Medicas

Aplicadas (CIMA), University of Navarra, Pamplona, Spain; 16COPPE-Computation Engineering &

Systems Post-graduation, Federal University of Rio de Janeiro, Rio de Janeiro(UFRJ), Brazil.

#EM, LH and AvdS-G share first authorship; $AO and VHJvdV share last authorship

of 24

Supplemental Methods

Sample processing and analysis. All samples were stained and acquired according to the EuroFlow

Standard Operating Procedure (SOP) and EuroFlow instrument settings.1 Appropriate instrument

performance and laboratory procedures were confirmed by results obtained in the EuroFlow quality

assessment rounds.2 Uniformity of data was evaluated as described in the Supplemental Results.

Gating of populations was performed using a standard Infinicyt profile (i.e. standardized population

tree and plots) in Infinicyt software (Cytognos, Salamanca, Spain). To facilitate uniform gating, detailed

gating instructions were discussed and subsequently described in a EuroFlow document, which

included all agreements regarding the annotations for EuroFlow databases, the gating strategy per profile, the profile for (sub)population definition, and the procedure for uploading the cases to the server (“Building the new EuroFlow database”, version 5.0, May

2014). After uploading cases to the EuroFlow server (Box A), cases were checked by a second

person. During checking, there were no major adaptations in gate settings, but in most cases the

leukemic population was somewhat further cleaned by additional removal of debris and doublets as

well as removal of unspecific APCH7 stainings.

Building of the EuroFlow ALOT database. All samples suspected for an acute leukemia and for

which an ALOT tube was run in one of the participating centers were eligible for inclusion in the study.

The vast majority of samples (>95%) used for building the database were bone marrow or peripheral

blood samples of de novo acute leukemia patients. Three samples had another origin (one

cerebrospinal fluid, two pleural effusions) and were included anyway since also for such samples (in

patients with possible isolated extramedullary disease) appropriate lineage assessment is crucial. The

same holds true for the 30 relapsed patients that were available. For last two sample categories we

assumed that the other tissue location or previous treatment might have an impact on the

immunophenotype of the leukemia, but that this will mainly reflect maturation/differentiation-related

markers rather than the lineage-defining markers which are present in the ALOT tube. Of importance,

none of the included relapsed patients received anti-CD19 targeted therapy (blinatumomab), that may

significantly alter CD19 expression. Therefore, it was decided that such samples could be included in

the database.

Since the aim of the database and algorithm is not to provide a diagnosis but rather to provide on an

individual cell basis the presence of typical B, T or myeloid cells, we built the database with typical B-

ALL, T-ALL and AML cases only. Consequently, atypical cases like mixed-phenotype acute leukemia

(MPAL), acute undifferentiated leukemia (AUL), and the rare blastic plasmacytoid dendrtic cell

neoplasm (BPDCN) were not included. In this way, such special and often complex cases will not be

recognized as typical by the database and will provide a warning to the user. For the same reason,

cases with obvious atypical expression of one or more markers (which could be due to technical

reasons or could reflect rare biology) were excluded from the final database during the quality check.

The resulting database therefore contains typical B-ALL, T-ALL and AML cases, allowing easy

identification of atypical or rare events present in a sample.

of 24

For building the database, all cases available (in Box C) were used, with the exception of cases from

Salamanca and Rotterdam. In this way, we added an additional level of difficulty for the algorithm

reflecting future usage, that is testing the database using data obtained in centers not contributing to

the database itself. Salamanca and Rotterdam cases were selected since they had sufficiently large

numbers and since they covered both pediatric and adult patients, thereby reflecting the broad range

of acute leukemia patients. A possible disadvantage of this approach is that the database contained

relatively low numbers of adult patients. Although differences in biology between pediatric and adult

patients are obvious, we assumed that these differences mainly reflect maturation/differentiation-

related markers rather than the lineage-defining markers which are present in the ALOT tube. We

therefore decided to use the advantage of our approach and accept the possible disadvantage.

Database-guided analysis (Compass tool). A PCA-based tool was developed to prospectively

classify individual leukemic cells from new patients in different categories, based on the sequential or

simultaneous comparison with cells from the AL patients in the database. Cells were classified as

follows:1. typical cells, related to single lineage cells (T, B or Myeloid) : the patient’s AL cells were

plotted against all three database groups simultaneously. The database group with most patient’s

cells within its 2SD curves (2SD boundaries derived from all cases in the database in a weighted

manner) was subsequently plotted in a separate unsupervised and balanced APS (PC1 vs PC2)

plot together with the patient’s AL cells; all patient’s AL cells that fell within the two SD curves of

the reference group were defined as typical (T, B, or Myeloid). All remaining cells were plotted

versus the other two reference groups separately; all patient’s AL cells that fell within the two SD

curves of one of these latter two reference groups were also defined as typical (T, B, or Myeloid).

Of note, once classified, cells are always excluded from the subsequent comparisons. They were

represented in the Compass tool as a single-color pointer directed to the appropriate disease

category associated with the percentage of the corresponding cells.

2. mixed cells: these mixed cells are positioned in areas of mixed potential in the supervised and

balanced APS plots. First, the remaining patient’s AL cells (after exclusion of all typical cells) are

plotted in an APS with all three reference groups and it is determined within which 2SD curves of

the reference groups most patient’s cells are positioned (dominant group) and towards which

other lineage these cells are moving (mixing group). If the percentage is ≥1%, the selected two

groups (dominant and mixed) are simultaneously projected in a new balanced and supervised

PCA comparison together with the patient’s AL cells. In this comparison, the events that are

outside 2 SD of the defined dominant reference group and towards the other group will be

considered as mixed. If several types of mixed phenotypes events are detected, the order of

classification will be based on the percentage of cells in each defined area in the comparison that

includes all three database groups. These cells were represented in the Compass tool as a dual-

color arrow directed towards the mixed component, associated with the percentage of cells.

3. transitional cells: This category includes patient’s AL cells that were not classified as typical T, B

or Myeloid cells, neither as mixed lineage cells, and that remain within the 2SD of the dominant

group, once one or more groups of mixed cells have been identified in the previous step. Thus,

of 24

there are no transitional cells in the absence of mixed cells. These cells were represented in the

Compass tool as a joint section (sector) in between typical and mixed components.

4. triphenotypic cells: triphenotypic cells are positioned outside the 2SD areas of the three groups

in all pair-wise comparisons and are positioned inside the square region in between the APS plot

with all three groups represented. These cells are represented as a grey circle whose diameter is

proportional to their percentage.

5. immature cells: cells that are positioned below a pre-defined zone (defined based on lines that

touch the lower parts of the 2SD curves of two groups). These cells are represented by a

percentage (%i) in the middle of the Compass tool.

6. unclassified cells: all events that did not fulfill any of the criteria for any of the above described

types of phenotype. These cells are not represented in the Compass tool.

Examples of the database-guided analysis are shown in Figures S4 , S5, S6, S7, S8, S9, and S10.

Local review of cases. Identification numbers of cases that were not classified as typical by the

database-guided analysis were sent to the local laboratory that enrolled the particular patient for

review. The local laboratory was requested to provide additional information with respect to the

complete immunophenotype (i.e. results from the classification panel(s)), additional genetic data

(particularly presence of BCR-ABL or MLL rearrangements), to review the original ALOT data, and to

confirm or revise the final diagnosis. Of note, the local laboratory was not aware of the results of the

database-guided analysis. Data were returned for central analysis and complex cases were discussed

by at least two independent persons until consensus was reached.

Statistical analyses. The probability of selecting the wrong antibody panel and make the wrong

diagnosis by using only ALOT (expressed as p-values) and their corresponding 95% confidence

intervals, as well as the required sample sizes for a p-value <0.05 or <0.08, were calculated based on

the number of cases evaluated and the actual mistakes made for all cases and per individual WHO

diagnostic subgroup using a proportional statistical test (MATLAB software, Mathworks, Natick, MA).

of 24

Supplemental Results

Uniformity in data between centers. All centers used the standardized EuroFlow protocols and

instrument settings.1, 3 Quality of standardization of acquired data was tested for each center twice a

year in a dedicated EuroFlow QA program. 2 Ninety-eight QA results reached score above 70% that is

considered acceptable (eighty-two were considered perfect) in 2015-2016. All issues found were

discussed and technicians were retrained if necessary.

To confirm the uniformity in data, Median Fluorescent Intensity values of the various markers were

evaluated on remaining normal cells present in the samples (CD19 and CyCD79a on B-cells, SmCD3,

CyCD3 and CD7 on T-cells, CD45 on lymphocytes, MPO on granulocytes) and compared between the

participating centers. As shown in Supplemental Figure S2, data were highly comparable between

centers. For all seven evaluated markers, the inter-laboratory variability (standard deviation of the

mean MFI values of the participating centers) was less than the intra-laboratory variability (standard

deviation of MFI values of measured samples) (Supplemental Figure S3).

Pediatric versus adult cases. Since the database was built using relatively few adult cases, we

evaluated the results from the validation cohort separately for pediatric and adult patients as well.

Results are summarized in Supplemental Table 3. the distribution of cases over the different

categories was comparable between the pediatric and adult cases. The two cases misclassified by the

database were both adult patients.

Agreement between ALOT and final diagnosis. Based on the validation series the p-value

(probability of error; in between brackets the 95% confidence interval) for the total cases and for each

WHO diagnostic subgroup separately are: total cases: 0.046 (0.022-0.070); AML: 0.046 (0.028-0.064);

BCP-ALL: 0.032 (<0.001-0.070); T-ALL: 0.085 (<0.001-0.216); MPAL: 0.049 (0.033-0.065). Based on

the misclassifications found, the estimated number of cases required for an error of <5% (p-value

<0.05) and <8% (p-value <0.08) for all cases together and per WHO diagnostic subgroup would be:

total cases: 68 / 27 cases; AML: 68 / 27 cases; BCP-ALL: 48 / 19 cases; T-ALL: 120 / 47 cases;

MPAL: 73 / 29 cases, respectively.

of 24

References

1. Kalina T, Flores-Montero J, van der Velden VH, Martin-Ayuso M, Bottcher S, Ritgen M, et al.

EuroFlow standardization of flow cytometer instrument settings and immunophenotyping

protocols. Leukemia 2012 Sep; 26(9): 1986-2010.

2. Kalina T, Flores-Montero J, Lecrevisse Q, Pedreira CE, van der Velden VH, Novakova M, et

al. Quality assessment program for EuroFlow protocols: summary results of four-year (2010-

2013) quality assurance rounds. Cytometry A 2015 Feb; 87(2): 145-156.

3. van Dongen JJ, Lhermitte L, Bottcher S, Almeida J, van der Velden VH, Flores-Montero J, et

al. EuroFlow antibody panels for standardized n-dimensional flow cytometric

immunophenotyping of normal, reactive and malignant leukocytes. Leukemia 2012 Sep; 26(9): 1908-1975.

of 24

Supplemental Table 1. Characteristics of patients used for constructing the EuroFlow ALOT database.

Diseasecategory

IncludedCases

ExcludedCases

T-ALL (n) 145 20Age (years) Median (min-max) 12 (0-69) 13 (0-66)

Gender (n)M 113 12F 32 8

NA 0 0

Sample type (n)

Peripheral blood 40 6Bone marrow 103 14

NAa 2 (PE)b 0

BCP-ALL (n) 377 45Age (years) Median (min-max) 4 (0-84) 8 (1-63)

Gender (n)M 213 28F 162 17

NA 2 0

Sample type (n)


NA 0 1 (1 CSF)c

AML (n) 134 26Age (years) Median (min-max) 10 (0-80) 10 (0-84)

Gender (n)M 63 14F 71 12

NA 0 0

Sample type (n)


NA 0 0

Total (n) 656 91Age (years) Median (min-max) 6 (0-84) 10 (0-84)

Gender (n)M 389 54F 265 37

NA 2 0

Sample type (n)Peripheral blood 68 14

Bone marrow 585 76NA 2 (2 PE) 1 (1 CSF)c

a NA: not available; b PE: pleural effusion; c CSF: cerebrospinal fluid.

of 24

Supplemental Table 2. Reasons for excluding cases from the database*.

Reason for exclusion Number of cases (n)

Atypical CD19 staining on blasts cells 2

Atypical CD45 staining on blasts cells 1

Atypical CD7 on staining blasts cells 5

Atypical CyCD3 staining on blasts cells 17

Atypical CyMPO staining on blasts cells 10

Atypical SmCD3 staining on blasts cells 9

Atypical CD45 staining on mature lymphocytes 17

Atypical SmCD3 staining on mature lymphocytes 2

Two atypical markers 13

Three Atypical markers 3

Total 81

*For illustrating examples, see also Supplemental Figure 11.

of 24

Supplemental Table 3. Results of database-guided analysis for pediatric and adult cases separated. *

Category Pediatric cases Adult cases Total

Typical T, B, or Myeloid 162 (65%) 373 (70%) 535 (68%)

Typical T, B, or Myeloid with

mixed populations <10%

37 (15%) 79 (15%)a 116 (15%)a

Typical T, B, or Myeloid with

mixed populations ≥10%

15 (6%) 39 (7%) 54 (7%)

Atypical 3 (1%) 6 (1%) 9 (1%)

Mixed lineage 31 (13%) 38 (7%) 69 (9%)

Total 248 535 783

*Data reflect number of patients; the percentages refer to the total number of pediatric, adults or all

patients. a Two cases were finally classified as MPAL and therefore were wrongly directed to a single

classification panel by the database-guided analysis. This represents 0.5% of all adult patients and

0.3% of all cases. Consequently, the database-guided analysis directs to the right classification panel

in 100% of pediatric patients, and over 99% in adult patients.

of 24

Supplemental Table 4 to 8:

Please see separate excel files with details about all cases.

of 24

Supplemental Figures

Figure S1. Schematic overview of the procedure designed and used for the construction of the EuroFlow ALOT database. Required annotations (date of acquisition, type of coagulant used for

preservation of sample, type of flow cytometer, age at diagnosis, gender, final classification according

to WHO2008 (T-ALL, BCP-ALL, AML, MPAL, AUL, other), disease phase (diagnosis, relapse), WBC

(optional)) were reported in an excel sheet

of 24

Figure S2. Comparability of MFI values of individual markers expressed by remaining normal cells. Median Fluorescence intensity (MFI) values were determined on remaining normal cells present

in the samples (CD19 (A) and CyCD79a (B) on B-cells, SmCD3 (C), CyCD3 (D) and CD7 (E) on T-

cells, CD45 (F) on lymphocytes, MPO (G) on granulocytes) and compared between the participating

centers. Number of evaluated samples for B cells, T cells and granulocytes, respectively, are: HAG:

358, 361, 330; KSA: 107, 107, 73; MON: 11, 8, 12; PRA: 89, 89, 88; RIO: 27, 27, 27; ZAB: 159, 156,

154; LIS: 16, 16, 12; PAR: 75, 75, 72’; PAM: 9, 8, 3; ROT: 480, 507, 373; SAL: 57, 59, 31.

of 24

Figure S3. Variability in MFI values of individual markers expressed by remaining normal cells. Median Fluorescence intensity (MFI) values were determined on remaining normal cells present in the

samples. Per center, the standard deviation of the obtained results was determined (intra-laboratory

variation). In addition, per marker the standard deviation was determined of the mean values of the

individual centers. Number of evaluated samples for B-cell markers (CD19 and CyCD79a), T-cell

markers (SmCD3, CyCD3, CD7 and CD45), and MPO, respectively, are: HAG: 358, 361, 330; KSA:

107, 107, 73; MON: 11, 8, 12; PRA: 89, 89, 88; RIO: 27, 27, 27; ZAB: 159, 156, 154; LIS: 16, 16, 12;

PAR: 75, 75, 72’; PAM: 9, 8, 3; ROT: 480, 507, 373; SAL: 57, 59, 31.

of 24

A.

B.

Figure S4. Example of database-guided analysis of a T-ALL (case 1). A. Database-guided analysis and definitions of lineage of single leukemic cell events. Virtually all leukemic cells fell within the 2SD curves of the T group already in the first step, and therefore immunophenotypes defined as typical T. Subsequent steps were not needed (<1% of events). B. Classical dot plots of the same patient show a clear T-ALL immunophenotype. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue) and granulocytes (purple) are shown as well.

of 24

A.

B.

Figure S5. Example of database-guided analysis of a BCP-ALL (case 2). A. Database-guided analysis and definitions of lineage of single leukemic cell events. Virtually all cells fell within the 2SD curves of the B group and therefore they were defined as typical B. Subsequent steps were not needed (<1% of events). B. Classical dot plots of the same patient show a clear BCP-ALL immunophenotype. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue), granulocytes (purple) and monocytes (orange) are shown as well.

of 24

A.

B.

Figure S6. Example of database-guided analysis of an AML (case 3). A. Database-guided analysis and definitions of lineage of single leukemic cell events. 90% of events fell within the 2SD curves of the Myeloid group and therefore they were defined as typical Myeloid. The remaining cells were evaluated in pair-wise comparisons, and no mixed cells were observed. Remaining events were non-classified . B. Classical dot plots of the same patient show a clear AML immunophenotype. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue) and immature erythroid cells (brown) are shown as well.

of 24

A.

B.

Figure S7. Example of database-guided analysis of an infant BCP-ALL (case 4; MLL rearranged). A. Database-guided analysis and definitions of lineage of single leukemic cell events showed only 20% of events within the 2SD curves of the B group (20% cells defined as typical B). The remaining cells were in-between Myeloid and B, Myeloid being the dominant group. When comparing the case with the B and Myeloid group, cells outside the 2SD of the Myeloid group were defined as mixed M/B cells, and the remaining cells within the 2SD were defined as transitional cells (BM/B). B. Classical dot plots of the same patient show a BCP-ALL immunophenotype with atypical low expression of CyCD79a and large scatter. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue), and granulocytes (purple) are shown as well.

of 24

A.

B.

Figure S8. Example of database-guided analysis of an MPAL (case 6). See legend on next page.

of 24

C.

D.

Figure S8. Example of database-guided analysis of an MPAL (case 5). A. Database-guided analysis and definitions of lineage of single leukemic cell events. None of the events fell within the 2SD curves of the three groups. Evaluation of mixed cells shows many cells outside all groups; the cells are therefore evaluated in group-by-group comparisons shown in B. Cells outside all three groups and within the defined MTB area (dashed lines) were defined as MTB. C. Remaining cells were again evaluated for mixed cells, showing Myeloid as dominant group and B as potential partner. Cells within the 2SD curves of Myeloid were defined mixed M/B cells, those inside the 2SD curves transitional cells (MM/B). D. Classical dot plots of the same patient show CD19+/CyCD79a+/ CD34+/CyMPO-/CD7+/weak CyCD3+ cells; the patient was diagnosed as MPAL. Additional classification protocols showed TdTweak+/ CD10-/ CD123+/ CD13-/ CD33-/ CD117-/ CD99weak+/ CD2+/ CD5weak+/ CD1a-/ TCR-/ CD4-/ CD8-/ CD45RA-/ HLADR-/ CD56partially+. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue), granulocytes (purple) and monocytes (orange) are shown as well.

of 24

A.

B.

Figure S9. Example of database-guided analysis of an MPAL (case 7). See legend on next page.

of 24

C.

Figure S9. Example of database-guided analysis of an MPAL (case 6). A. Database-guided analysis and definitions of lineage of single leukemic cell events. 30% of cells was defined as typical T. The remaining events could be mixed T/B or MTB. B. The mixed cells did not fulfill the criteria for MTB and hence were classified as T/B C. Classical dot plots of the same patient show CyMPO-/CD7+/CyCD3+/CyCD79a weak +/CD19weak+ cells, compatible with MPAL (T/B lineage). Additional classification protocols showed CD45dim/ TdT+/ CD10weak+/ CD20-/ CD38+/ CyIgM-/ SmIgM-/ CD22-/ CD24-/ CD81+/ CD21-/ CD123weak+/ CD66c-/ CD9-/ CD13 partially weak +/ CD33weak+/ CD15-/ CD117 partially weak+/ CD99+/ CD2-/ CD5weak+/ CD1a-/ TCR-/ CD4-/ CD8-/ CD45RA+/ HLADRweak+/ CD56-. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue), granulocytes (purple), and immature erythroid cells (brown) are shown as well.

of 24

A.

B.

Figure S10. Example of database-guided analysis of an MPAL (case 8). A. Database-guided analysis and definitions of lineage of single leukemic cell events. 54% of cells was defined as typical T and 43% of cells was defined as typical Myeloid. The remaining events do not show characteristics of mixed cells. B. Classical dot plots of the same patient show CyMPO+/CD7+/CyCD3+ cells, compatible with MPAL (T/Myeloid). Additional classification protocols showed CD45weak+/ CD1-/ CD2+/ CyCD3weak+/ CD3-/ CD4-/ CD5-/ CD7+/ CD8-/ CD10-/ CD13weak+/ CD19-/ CD33-/ CD34weak+/ CD56-/ CD117+/ CD99+/ HLA-DRweak+/ CyMPOpartially+/ TdT+/ CyCD79a-. Leukemic cells: red. Normal T-cells (dark blue), NK cells (middle blue), mature B-cells (green), eosinophils (light blue) and granulocytes (purple) are shown as well.

of 24

Figure S11. Identification of cases with atypical expression during quality assessment of cases to be included in the ALOT database. In this example, BCP-ALL cases (n=422) were evaluated for

the expression of A. CD3, B. CD45, C. CyMPO or D. CyCD3 on either the leukemic BCP-ALL cells or

normal mature lymphocytes, allowing identification of cases with atypical/unexpected expression.

Cases marked by the red circle showed clearly atypical expression of the involved marker as

compared to the vast majority of samples (in the grey box, representing the two SD borders). These

differences might be due to technical errors or due to biological issues, but clearly disqualifies these

cases as typical. The different symbols reflect data derived from different laboratories.

of 24

A.

B.

Figure S12. PCA of the three reference groups. A. PCA of the 377 BCP-ALL (green), 145 T-ALL

(blue) and 134 AML cases (red) shows clear and separate clusters of the three categories. Dots

represent the median values of individual patients, the dotted line represents the 1SD curve of the

group and the solid line represents the 2SD curve. B. Contribution of each parameter to the first (PC1,

x-axis) or second (PC2, y-axis) principal component reflected as percent values.

of 24

1 SD

2 SD

media.nature.com · web viewautomated database-guided expert-supervised orientation for...

Documents