AMP Benchmark Datasets

Curated datasets constituting a comprehensive benchmark compiled from ABP entries across the current AMP database ecosystem, training datasets of existing prediction tools, and non-AMP entries sourced from UniProt, designed to enable systematic and reproducible evaluation of AMP prediction tools. These datasets were used to assess prediction tools reviewed in our publication (preprint: https://doi.org/10.64898/2026.05.19.726290), with evaluations executed through a Nextflow pipeline. Code for this and additional analyses described in the publication is available at https://github.com/Maximo-7/amp_review.git. This Zenodo deposit contains three items: Complete dataset: all curated entries with full annotation. Evaluation subset: the subset used for benchmarking prediction tools. Column dictionary: definitions for all fields in the complete dataset.