Background: B cell acute lymphoblastic leukemia (B-ALL) is the most common malignancy in children. Although overall survival (OS) for pediatric B-ALL is around 85% at 5 years, ultimately more than 20% of affected children succumb to relapse. Leukemia stem cells (LSCs) which cause relapse and chemo-resistance are believed to relate to patient prognosis. We aim to use LSC gene signature to develop an easy way to predict patient risk at diagnosis.
Methods: Publicly available LSC RNA-seq for B-ALL was obtained from GEO and analysed using RNAsik and edgeR. Training data was obtained directly from TARGET website. Test datasets were microarray from GEO and TARGET website. LASSO regression was done on training data with 10 folds cross validation (CV). CV was performed 100 times randomly to select top 3 models with most occurrence. All models generated were then tested in all three test datasets for validation based on hazard ratio and p value. Survival analysis was based on Cox Proportional hazard model.
Result: Differentially expressed genes from LSCs were enriched in pathways related to immune response and cell cycle arrest. Genes upregulated in LSCs with significant adverse survival impact were selected for LASSO regression. The final model is 0.065*S100A10+0.051*ZMAT3+0.017*PSAT1+0.108*RIMS3+0.01*LRRC25+0.015*H1FX+0.04*TSPO+0.029*NID2+0.014*CCDC69. It was validated in training data and all three test data including 2 paediatric and 1 adult B-ALL from different platforms. Moreover, it not only worked in full dataset, but also in a subset of patients with uninformative cytogenetics.
Conclusion: We are able to develop a 9-gene score to fast calculate the risk of patient. The score is agnostic to platform, patient age and uninformative cytogenetics.