Classification of Pediatric Acute Lymphoblastic Leukemia by Gene Expression Profiling

Supplemental Information for:

Classification OF Pediatric Acute Lymphoblastic Leukemia by Gene Expression Profiling

Mary E. Ross1, Xiaodong Zhou2, Guangchun Song2, Sheila A. Shurtleff2, Kevin Girtman2, W. Kent Williams2, Hsi-Che Liu2, Rami Mahfouz2, Susana C. Raimondi2, Noel Lenny2, Anami Patel2, & James R. Downing2,*

Table of Contents

Section I: Patient Dataset

Diagnostic ALL samples used for class prediction

Subgroup distribution of ALL cases

Section II: Methods

Hybridization of microarrays

Statistical methods

Section III: Genetic Subtype Discriminating Genes

Top 100 chi-square probe sets selected for BCR-ABL decision tree format

Top 100 chi-square probe sets selected for E2A-PBX1 decision tree format

Top 100 chi-square probe sets selected for Hyperdiploid >50 decision tree format

Top 100 chi-square probe sets selected for MLL decision tree format

Top 100 chi-square probe sets selected for T-ALL decision tree format

Top 100 chi-square probe sets selected for TEL-AML1 decision tree format

Top 100 chi-square probe sets selected for BCR-ABL parallel format

Top 100 chi-square probe sets selected for E2A-PBX1 parallel format

Top 100 chi-square probe sets selected for Hyperdiploid >50 parallel format

Top 100 chi-square probe sets selected for MLL parallel format

Top 100 chi-square probe sets selected for T-ALL parallel format

Top 100 chi-square probe sets selected for TEL-AML1 parallel format

Section IV: Diagnostic Accuracy

Training and test set results

Cross comparison of supervised learning algorithms

Section V: Comparison of Expression Profiles and Real-time PCR (Taqman)

Section VI: References

I: Patient Dataset

132 cases of pediatric ALL were selected from the original 327 diagnostic bone marrow aspirates1 to reanalyze on the higher density U133A and B microarrays. The selection of cases was based on having sufficient numbers of each subtype to build accurate class predictions, rather than reflecting the actual frequency of these groups in the pediatric population. The list of samples that were used in this reanalysis (Table S1), as well as the subtype distribution (Table S2) are shown below.

Table Key:The nomenclature used in this paper is identical to that used in Yeoh et. al.,1 and thus should facilitate cross comparisons between the datasets. The nomenclature indicates disease status at the time of the initial study and has not been updated as this dataset was not selected to address the issue of outcome. No analysis has been performed in this study to identify expression profiles associated with outcome.

Subtype Name-C# Dx Sample of patient in CCR
Subtype Name-R# Dx Sample of patient who developed a hematologic relapse
Subtype Name-# Dx Sample used for subgroup classification only
Subtype Name-2M# Dx Sample of patient who later developed 2nd AML
Subtype Name-N Dx Sample in novel group

MLL

II: Methods

Hybridization of microarrays

Hybridization solutions from our previous U95A study had been stored at -80oC since their initial use. These solutions were thawed at 45oC, then microcentrifuged for 2 minutes to remove any insoluble material from the mixture. The hybridization solutions were added to U133A chips and allowed to hybridize for 16 hours at 45oC. At the end of the incubation period, the hybridization solution was removed from each U133A chip and refrozen. Subsequently, the hybridizations were thawed and hybridized to the U133B chip.

A non-stringent wash buffer (6X SSPE, 0.01% Tween 20) was added to each chip cassette after the hybridization solution was removed and the cassette allowed to equilibrate to room temperature. The microarray cassettes were then placed on the fluidics station and the antibody amplification protocol performed. The arrays were washed at 25oC with the non-stringent buffer followed by a more stringent wash at 50oC with 100 mM MES, 0.1M NaCl2, 0.01% Tween 20. The arrays were then stained with Streptavidin Phycoerythrin (SAPE, Molecular Probes, Eugene, OR) for 10 minutes at 25oC. Following another non-stringent wash, the arrays were hybridized for 10 minutes at 25oC with an antibody solution (100 mM MES, 1 M [Na+], 0.05% Tween 20, 2 mg/ml BSA, 0.1 mg/ml goat IgG, and 3 g/ml biotinylated antibody). This solution was removed and the cassettes restained with the SAPE solution.

Arrays were scanned on a laser confocal scanner (Agilent, Palo Alto, CA) and then analyzed with Affymetrix Microarray Suite 5.0 (MAS 5.0). Detection values (present, marginal or absent) were determined by default parameters, and signal values were scaled by global methods to a target value of 500. After completing the scans, the arrays were visually inspected for defects and Affymetrix internal controls were utilized to monitor the success of hybridization, washing, and staining procedures.

Statistical methods

The chi-square metric and the k-NN and ANN supervised learning algorithms have been previously described. For more information see The SVM supervised learning algorithm that was used in this study is available as part of the software package Rv 1.6.0.

To determine the performance of each model using ANN, a confidence threshold was built for each diagnostic subtype utilizing a modification of the method described by Khan et al.2 Models were built based on a decision tree format where each level of the decision tree contains only two possible distinctions – class and non-class (for example, T verses non-T). At each level, using only samples in the training set, 3 ANN models were built by 3-fold cross validation. The training set samples were then shuffled and 3 additional ANN models were built. This model building process was repeated for a total of 100 times at each step of the decision tree. Then an empirical probability distribution for the ANN output node value was built only for subtype under study, for example, T-ALL at the first step of the decision tree. Only nodal values greater than 0.5 for each subtype were included. For each individual sample in the training set, the 100 validation subtype node values were averaged and compared to threshold. Individual samples were assigned to the subtype under study only when its average subtype nodal value was greater than the 95% confidence threshold. For samples in the test set, subtype nodal values are averaged from all models generated in the 3-fold cross validation. A sample is assigned to the class under study when the average subtype nodal value is greater than the 95% confidence level defined on the training set. A sample not assigned to the subtype will progress to the next level of the decision tree, where the entire process is repeated.

1

III: Genetic Subtype Discriminating Genes

The following tables contain a listing of the top 100 probe sets for each diagnostic subtypes ranked by their chi-square value (Tables S3-S8). Each table contains the Affymetrix U133 series probe set number, a gene description, gene symbol, chromosomal location, and primary GenBank reference. Chi-square values were calculated utilizing only the samples in the training set in a differential diagnosis decision tree format as discussed in the text and illustrated in Figure 2 (Tables S3-S8) or by a parallel approach (Tables S9-S14). The calculation of the fold change was done in a parallel format using the total data set and comparing the mean signal value in the class versus the mean signal value in the non-class. The last column indicates whether this gene had previously been identified as a class discriminator using the U95Av2 data (old) or identified as a class discriminator only using the U133 data (new) (Tables S3-S8).

Table S3. Top 100 chi-square probe sets selected for BCR-ABL in decision tree format

Table S4. Top 100 chi-square probe sets selected for E2A-PBX1 in decision tree format

Table S5. Top 100 chi-square probe sets selected for Hyperdiploid >50 in decision tree format