Benchmarking Peptide-Protein Affinity Prediction Across Peptide and Target Shifts
Testing whether AI can predict drug-like peptide binding when it sees new targets
Researchers tested ten different methods for predicting how strongly peptides bind to proteins, using 11,349 binding measurements split into three realistic scenarios: recognizing patterns in similar peptides, making predictions within known targets, and predicting binding to entirely new target proteins. The best method changed depending on the scenario—fingerprints worked best for familiar targets, but a model called HELM-BERT performed better when targets were completely new. Performance dropped significantly (from 0.67 to 0.53 correlation) when moving from known to unknown targets, showing that current methods struggle with genuine generalization.
Drug discovery relies on predicting how experimental molecules will bind their targets before expensive testing begins. This work exposes a critical gap: most peptide-binding models are benchmarked in unrealistic conditions and would fail in the real world, where researchers must predict binding to new disease targets they've never seen before. The findings demand that future benchmarks test what actually matters—whether methods can handle unfamiliar targets—and reveal which representation strategies hold up under that harder test.