Integrated Multi-Optosis Model for Pan-Cancer Candidate Biomarker and Therapy Target Discovery

Abstract

Regulated cell death (RCD) is essential for maintaining tissue homeostasis and controlling cellular stress responses, which are crucial in both cancer suppression, progression, and treatment response across multiple tumor types. RCD processes function in interconnected and overlapping networks. Although few individual RCD forms have been extensively studied, there is a lack of comprehensive approaches integrating multiple RCD forms, limiting the potential for holistic biomarker discovery. Our study addresses this gap by integrating 25 RCD forms into a multi-optosis model that correlates multi-omic and phenotypic data across 33 cancer types using a multilayer approach, enabling the discovery of candidate biomarkers with genome-wide significance. The model was developed using multi-omic data, including tumor and non-tumor tissue samples, from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) databases, accessed via UCSCXena and UCSCXenaShiny. 9,185 tumor samples from TCGA and 7,429 non-tumor tissue samples from GTEx were analyzed. We queried 5,913 genes associated with 25 RCD forms encompassing 62,090 transcript isoforms, 882 mature miRNAs, and 239 cancer-associated proteins and protein modifications. Seven omic features were examined: mutations, copy number variations (CNV), CpG methylation, protein array, mRNA, miRNA, and transcript isoform expression, all correlated with seven clinical phenotypic outcomes: tumor mutation burden (TMB), microsatellite instability (MSI), tumor stemness (TSM), hazard ratio, prognostic survival, tumor microenvironment (TME), and tumor immune cell infiltration (TIL). Over 27 million pairwise correlations were performed between these phenotypic features and multi-omic data. Our findings reveal 44,641 multi-omic signatures, comprising both unique signatures specific to individual RCD forms and overlapping signatures across multiple RCD types, highlighting significant correlations between the omic features and phenotypic outcomes. Each signature received a structured alphanumeric identifier, encoding its biological context, including multi-omic and phenotypic correlations and cancer specificity, providing a systematic approach to categorizing and tracing these associations. Apoptosis-related genes were prevalent across most multi-omic signatures, reaffirming the pivotal role of apoptosis partners among diverse RCD pathways in cancer. Our analysis revealed a prevalent occurrence of isoform-specific signatures, where transcript isoforms originating from the same gene exhibited distinct phenotypic correlations. This finding highlights the intricate roles of alternative splicing and promoter usage in regulating cancer. For instance, in multi-transcript genes like MAPK10, specific isoforms were associated with unique phenotypic outcomes, emphasizing the importance of isoform-level resolution for understanding cancer progression and therapeutic response. Notably, rare exceptions were observed in multi-transcript genes such as COL1A1 and UMOD, where all isoforms consistently correlated with stemness indices in a cancer-type-specific manner, showing a coordinated regulatory function at the gene level rather than the isoform level. This finding underscores the potential for specific genes to exert unified effects on tumor biology, which may be crucial for maintaining stemness profiles across cancer types. Significantly, 879 of the identified RCD multi-omic signatures include chimeric antigen targets currently under clinical evaluation, emphasizing the translational potential of these multi-omic signatures in immunotherapy and precision oncology. This integrative framework, accessible through the CancerRCDShiny tool (https://cancerrcdshiny.shinyapps.io/cancerrcdshiny/), provides a powerful resource for advancing candidate biomarker discovery and identifying actionable therapeutic targets across cancer types with specific applications in immunotherapy.