Bimodality in pan-cancer proteomics reveals new opportunities for biomarker discovery

Abstract

Bimodal protein expression, characterized by the distribution of protein expression with two modes, is linked to phenotypic variation across various biological systems. Whereas previous studies focused on RNA expression data, we developed a bimodality model tailored for proteomics to enhance the identification of cancer-associated biomarkers and targets, facilitating precision oncology. We analyzed proteomics data from various cancer types and identified 2401 tumor-associated bimodal proteins. These proteins were evaluated for pathway enrichment, revealing significant associations with critical cancer pathways, such as metabolism of non-essential amino acids, interaction between the extracellular matrix and its receptors on the cell surface, and central carbon metabolism in cancer. Utilizing an AI-enhanced knowledge graph, we further delineated common patterns among pan-cancer tumor-associated bimodal proteins. A case study on the bimodal expression of TROP2 in colon adenocarcinoma highlighted upregulation of MYC and WNT/{beta}-catenin signaling pathways and down-regulation of inflammatory and interferon-related pathways in the TROP2-high group. The biological difference between TROP2-high and TROP2-low groups underscored its significance in determining cancer heterogeneity and differences in cancer vulnerability, which can inform treatment decisions. Our findings show the value of proteomics in uncovering novel biomarkers and advancing precision medicine, setting a precedent for further multi-omics integration and clinical validation.