Invariant Differential Expression Analysis Reveals Mechanism of Cancer Resistance to Cell Cycle Inhibitors

Abstract

Retinoblastoma (RB) is a good model to study drug resistance to cell-cycle inhibitors because it is driven by mutations in the core components of cell-cycle, i.e, Rb gene. However, there is limited gene expression dataset in RB which has major reproducibility issues. We have developed invariant differential expression analysis (iDEA) that improves the state of the art in differential expression analysis (DEA). iDEA uses strong Boolean implication relationships in a large diverse human dataset GSE119087 (n = 25,955) to filter the noisy differentially expressed genes (DEGs). iDEA was applied to RB datasets and a gene signature was computed that led to prediction and mechanism of drug sensitivity. The prediction was confirmed using drugs-sensitive/resistant RB cell-lines and mouse xenograft models using CDC25 inhibitor NSC663284. iDEA improved reproducibility of differential expression across diverse retina/RB cohorts and RB cell-lines with different drug sensitivity (Y79/Weri vs NCC). Pathway analysis revealed WNT/{beta}-catenin involved in distinguishing drug sensitivity to CDC25 inhibitor NSC663284. NSC663284 inhibited tumour cell proliferation in mouse xenograft model containing Y79 cells indicating novel therapeutic option in RB. Invariant differentially expressed genes (iDEGs) are robustly associated with outcome in diverse cancer datasets and supports for a fundamental mechanism of drug resistance.