ESCAPE: A Standardized Benchmark for Multilabel Antimicrobial Peptide Classification

Current research on AMP prediction faces a critical bottleneck due to fragmented, inconsistent, and limited-scale datasets that differ widely in format, annotation schema, and functional coverage, hindering the development of robust models and making difficult fair comparisons between approaches. To address this gap, we introduce ESCAPE (Expanded Standardized Collection for Antimicrobial Peptide Evaluation) dataset, a unified collection of antimicrobial peptides compiled from 27 public repositories, structured around four primary antimicrobial activities, along with an additional label indicating whether the peptide is antimicrobial or non-antimicrobial. ESCAPE integrates diverse functional profiles across peptides, capturing the full spectrum of antimicrobial activities and establishing a comprehensive benchmark for AMP prediction.