Employing serial Cox modeling for identification of pan-cancer biphasic genes

Abstract

Abstract Motivation: Some genes, termed 'biphasic genes', can transition between a preventive and promoting effect on disease recurrence during cancer progression. Identifying these genes poses substantial challenges for conventional statistical methods, such as Cox Proportional Hazard analysis. Addressing this issue, the present study introduces an algorithm to pinpoint biphasic genes in 17 TCGA cohorts of RNA sequencing high-throughput data. Moreover, the detected biphasic genes appear instrumental in biological processes essential for the adaptive responses in cancer progression. Results: This approach identified a total of 365 unique biphasic genes across 17 TCGA cohorts, high-lighting their essential roles in dynamically influencing progression-free interval lengths. Most genes displayed differential directions of effect across cohorts, possibly corresponding to their context-de-pendent nature. The gene set enrichment analysis not only unveiled diverse functional domains, from signaling transduction and cellular transport to proliferation and energy metabolism, but also hinted at various possible future research directions for elucidating the role of biphasic genes in cancer progres-sion and dynamic disease responses. Contact: frank.hause@uvic.cat Supplementary information: Supplementary data as well as all R code associated with the present analysis are available at https://github.com/DataScienceFH/BiphasicGenes-TCGA.