Abstract 5022: iS3RGs: Discriminating diverse medulloblastoma subtypes by leveraging heterogenous transcriptome data with batch effects

Abstract As the most common malignant pediatric brain cancer, medulloblastoma (MB) accounts for around 20% of all pediatric central nervous system (CNS) neoplasms. MB includes a complex array of distinct molecular subtypes characterized by morphological, molecular, and genetic alterations, mainly including SHH, WNT, Group3 and Group4. Although the overall cure rates for MB are around 70%, the survival rate varies significantly across different molecular subtypes. Conventional experimental methods for identifying MB subtypes are usually costly, time-consuming, and labor-intensive. Accurate identification of MB subtypes enables improved downstream risk stratification and tailored therapeutic treatment design. Previous studies indicate that identifying MB subtypes based on transcriptome data is an efficient and feasible way. However, their performance may be poor due to limited cohorts and severe batch effects if combining data in different sequencing platforms and various data sources. To overcome these limitations, we propose a novel computational approach based on intra-sample subtype-specific ranked genes (iS3RGs) analysis to effectively integrate heterogeneous transcriptomic data with batch effects for accurate identification of MB subtypes across diverse cohorts. For a baseline microarray dataset comprising 763 samples as training part, we specifically converted the absolute gene expression to the intra-sample rank values to capture reliable gene patterns by reducing bias from fluctuations and measurement errors. By comparing gene pairs within cohorts rather than across cohorts, we could avoid batch effects and successfully identified those gene pairs that significantly contributed to the identification of MB subtypes. Subsequently, we performed differential ranking analysis for iS3RGs identification. These genes were organized into gene ranking ratios and the subtype-specific gene ratios were filtered using Fisher’s exact test and LASSO regression, based on which 1035 gene ratios (corresponding to 738 unique genes) were identified. Then by testing on 14 independent datasets, our proposed method significantly outperformed state-of-the-art MB subtyping approaches with a median accuracy of 99.02%. We believe that our proposed approach has direct positive impacts on downstream MB risk stratification and tailored treatment design. We also expect that our proposed approach has great potential to address many biomedical problems with batch effects for various types of pediatric, adolescent, and young adult (AYA) cancers. Citation Format: Mengtao Sun, Jieqiong Wang, Shibiao Wan. iS3RGs: Discriminating diverse medulloblastoma subtypes by leveraging heterogenous transcriptome data with batch effects [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5022.