Related Experiment Video
Updated: May 22, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Machine learning integration of multi-omics data develops a sarcomatoid-related renal cell carcinoma score (SARS) for
Dingbang Liu1, Xiaozhi Xia1, Wei Wang1
1Department of Urology, Institute of Urology, West China Hospital, Sichuan University, Chengdu, China.
Background:
Clear cell renal cell carcinoma (ccRCC) exhibits significant molecular heterogeneity, which limits the benefit of immunotherapy and underscores the urgent need for robust predictive biomarkers. The highly aggressive sarcomatoid phenotype, with its distinct treatment responses, provides a promising biological foundation for model development. This study aimed to develop and validate a machine-learning-based multi-omics score derived from sarcomatoid-associated genes [sarcomatoid-associated renal cell carcinoma score (SARS)] for prognostic stratification and prediction of therapeutic efficacy across different treatment modalities in advanced ccRCC.
Methods:
A machine learning ensemble algorithm was applied to sarcomatoid-associated genes to develop the SARS. The model stratification power was validated for overall survival (OS) and progression-free survival (PFS) across IMmotion151, CheckMate025, and JAVELIN Renal 101 trials. Multi-omics analyses, including single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, and metabolic flux inference, were used to decipher the associated tumor biology, tumor microenvironment (TME), and metabolic reprogramming.
Results:
The SARS model robustly stratified OS and PFS in both the training and external validation cohorts. SARS effectively stratified patients into SARS-high and SARS-low groups across all cohorts for both OS and PFS, and stratified the efficacy of monotherapy and combination therapy. Multi-omics characterization revealed that SARS-high tumors exhibited enhanced proliferative capacity, metabolism reprogramming, and an immunosuppressive TME. We further identified DRAP1 as a core gene within the SARS signature. Spatial transcriptomics confirmed DRAP1's specific overexpression in malignant regions, and single-cell analyses indicated its critical role in intercellular communication and early tumor progression.
Conclusions:
The SARS model serves as a powerful and integrative biomarker for prognostic stratification and therapeutic guidance in advanced ccRCC. DRAP1 is a potential novel therapeutic target, while further studies are needed to elucidate its underlying mechanisms.