PROTEOMICS APPROACH FOR IDENTIFYING NOVEL CANCER BIOMARKERS AND THERAPEUTIC TARGETS

Abstract

Cancer remains one of the leading causes of mortality worldwide, and the search for reliable biomarkers and druggable targets continues to shape both clinical practice and translational research. Genomic approaches have delivered enormous insight over the past two decades, but genes tell only part of the story. Proteins are the actual effectors of cellular function, and their abundance, modifications, and interaction patterns often diverge from what transcript-level data would predict. This paper examines how modern proteomics, particularly mass spectrometry-based approaches enriched with machine learning analysis, can be applied to the discovery of novel cancer biomarkers and therapeutic targets. We designed an integrated discovery framework that combines label-free quantitative proteomics, phosphoproteomic enrichment, and a computational pipeline for differential abundance analysis, pathway mapping, and target prioritization. The framework was evaluated on a simulated multi-cohort study covering breast, lung, colorectal, and pancreatic cancers with matched tumor and adjacent normal tissue samples from 1,240 patients. Results show that the integrated approach identified 187 candidate biomarkers with area under the ROC curve above 0.85 for at least one cancer type, of which 42 were shared across two or more cancers and 23 corresponded to proteins with existing or emerging drug modalities. Machine learning classifiers trained on the top proteomic signatures achieved diagnostic accuracy of 91.4% and outperformed transcript- only classifiers by roughly 8 percentage points. We discuss trade-offs including sample preparation variability, cost, and the challenge of translating discovery-stage findings into validated clinical assays. Findings support proteomics as a mature and increasingly essential pillar of cancer biomarker discovery, particularly when combined with computational integration across post-