Abstract 6215: Systematic identification and clinical implications of glycogenes in cancer research: An integrative omics approach

Abstract

Abstract Glycosylation and its products are crucial in clinical routines but are overshadowed by a framework embedding genomic participants in cancer development. The glycosylation scheme varies between research groups, presenting clinical concerns that demand understanding in cancer research beyond the glycome. Our work aims (a) to enhance the canonical framework by identifying potential glycosylation-associated genes (glycogenes) and (b) to advance translational cancer study, offering a unified perspective on integrative omics and explainable machine learning with glycogene integration. The expansion of the glycosylation framework involved initial text mining from curated databases like OMIM, InterPro, and Reactome. Dysregulation of identified genes was affirmed across 29 TCGA solid cancers, encompassing over 7,000 patients with primary tumors. And healthy tissue data from GTEx facilitated tumor-normal comparisons. Subsequently, integrative omic analysis, involving transcriptome and methylome, was applied to confirm the stratification feasibility of glycogenes and to revealed cancer clusters within the UMAP space. After cluster discovery, profiling from survival analysis and drug response prediction highlighted clinical differences between clusters. Employing a survival forest model and explainable artificial intelligence (XAI), personalized assessments, including a 5-year survival predictor and risk factors for individuals, were developed. Our study verified over 3,000 glycogenes, expanding the conventional configuration of glycosylation and constituting 10-30% of differentially expressed genes between tumors and healthy tissues. Whether transcriptome or methylome alone, glycogenes demonstrated stratification ability. Advanced omic integration further discovered 16 cancer clusters out of 29 TCGA cancers, associated with anatomical source, morphology, and genomic similarity. Utilizing drug resources from GDSC, this specialized geneset exhibited drug predicting ability, displaying a distinct drug response profile among the 16 clusters. To finalize the clinical potential, a survival forest for 5-year survival model construction concluded forecasters with a time-dependent AUC (0.77 to 0.92 on average) for each cancer cluster. With XAI implementation, survival risk factors were developed and assessed for each individual locally. Here we identified over 3,000 glycogenes, elaborating the classic glycosylation structure and emphasizing genomic importance in cancer research. Glycogenes demonstrated translational potential, supported by diverse survival distributions and drug responses across cancer clusters discovered by the integrative omic approach. The integration of XAI provided a tailored perspective for implementing glycogenes in clinical cancer research. Citation Format: Yen-Hsieh Chen, Yuh-Shan Jou. Systematic identification and clinical implications of glycogenes in cancer research: An integrative omics approach [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6215.