Personalized risk stratification in colorectal cancer via PIANOS system
Abstract
Background: Colorectal cancer (CRC) is the third-most common malignancy and the second-leading cause of cancer-related death worldwide. Despite considerable progress in therapeutic approaches for CRC, patient outcomes remain unsatisfactory, with a general five-year survival rate of ~65%. A robust, accurate, and clinically actionable risk stratification system provides a foundation for further drug discovery to improve CRC patient outcomes.
Methods: We present PIANOS, a robust, platform-agnostic classifier for stratifying CRC patient risk. We analyzed data from 24 cohorts across ten countries comprising 5439 patients. We used the single-sample gene set enrichment analysis (ssGSEA) algorithm to compute enrichment scores for 22,596 pathways from the MSigDB database for each patient. PIANOS was developed using k-TSP and resampling algorithms, training with 364 patients.
Results: PIANOS stratified patients into high- or low-risk categories based on gene expression profiles, designating those scoring >17 as high-risk. Multivariate analysis for disease-free survival (DFS) revealed that PIANOS stratification operates independently of TNM staging and MSI status. DFS in high-risk patients was considerably shorter compared to those in the low-risk group across all examined cohorts. PIANOS suggests that low-risk patients may benefit more from chemotherapy and immunotherapy, whereas high-risk patients may exhibit activated angiogenic features.
Conclusion: This study underscores the potential of PIANOS to enhance personalized clinical decision-making in CRC through robust, platform-agnostic risk stratification that identifies distinct treatment sensitivities.
Du CAI
Postdoc
I focus on leveraging explainable AI and large foundation models to advance medical imaging and digital pathology in colorectal cancer research.
Hao-Ning QI
Postdoc
Cheng-Hang LI
Research Assistant
Chu-Ling HU
PhD Student
Bao-Wen GAI
PhD Student
Feng GAO
Professor
My research leverages AI and big data to improve diagnostics, prognostics, and ultimately, outcomes in cancer and other biomedical fields.