Personalized risk stratification in colorectal cancer via PIANOS system

Abstract

Abstract Current prognostic biomarkers for colorectal cancer (CRC) lack stability and generalizability across different cohorts and platforms, challenging precise patient stratification. Here, we introduce a Platform Independent and Normalization Free Single-sample Classifier (PIANOS), designed to refine treatment decisions by accurately categorizing patients with CRC into distinct risk groups. Developed using gene expression data from 562 patients and employing a rank-based k-Top Scoring Pairs (k-TSP) algorithm alongside resampling, PIANOS was rigorously validated in 15 cohorts comprising 3666 patients with CRC. It effectively differentiates high-risk from low-risk patients, outperforms 105 existing models, and demonstrates robust performance across technologies like microarrays and RNA sequencing. PIANOS-based stratification is validated as an independent predictor of disease-free survival. Moreover, PIANOS discriminates treatment responses across risk categories, with high-risk patients showing increased sensitivity to bevacizumab and low-risk patients exhibiting enhanced responsiveness to chemotherapy and immunotherapy. This study reports significant advancements in supporting clinical decision-making for CRC and provides a reliable framework for optimizing patient treatment strategies.