AI-Identified Biomarkers Require Rigorous Validation Beyond Initial Accuracy for Clinical Utility
A recent review published in the journal Signal Transduction and Targeted Therapy highlights that while artificial intelligence (AI) can accurately identify promising disease signals or biomarkers from various data sources, their clinical utility depends on rigorous validation beyond initial predictive power. The review emphasizes that many AI-derived biomarker candidates, despite performing well in initial studies, often fail to be reproduced in independent cohorts or improve clinical decisions. The authors stress the need for biomarkers to be reproducible, biologically relevant, clinically applicable, and prospectively validated. AI's ability to integrate multi-scale data, including genomic, proteomic, imaging, and digital information, is powerful, but challenges such as bias, confounding, data heterogeneity, overfitting, and weak external validation can limit their translation into patient care. The review calls for a development pathway that spans discovery, external validation, testing under different...
