iDAOP antioxidant peptide benchmark: confirmatory similarity-aware analysis and reproducibility archive

This record contains machine-readable materials supporting the confirmatory similarity-aware analysis of the iDAOP antioxidant peptide benchmark. The archive includes benchmark-structure diagnostics; candidate sequence-relatedness grouping outputs; the primary C2 conservative sequence-relatedness grouping; grouped three-fold assignments across five prespecified seeds (42–46); row-level out-of-fold predictions and pooled performance metrics for amino-acid-composition (AAC) and frozen ESM-2 representations with logistic regression and linear SVM classifiers; a C3 k-mer grouping sensitivity analysis; train–test relatedness checks; LinearSVC convergence verification; analysis scripts; and software-environment information. The confirmatory core dataset comprises 899 positive-core and 29 core-negative sequences. Three sensitivity-negative sequences are evaluated separately in a test-only sensitivity analysis, while 28 challenge-only sequences are excluded from model fitting and headline performance estimates. The archive provides the machine-readable materials required to inspect the sequence-grouping structure, reproduce the confirmatory evaluation workflow, verify the reported performance summaries, and audit the associated quality-control checks. Historical benchmark materials and analyses are preserved separately in the iDAOP v1.0.5 release (Zenodo DOI: 10.5281/zenodo.22107311). This confirmatory-analysis archive is a separate supporting research artifact and is not a replacement for, or new version of, that historical release. Version 1.0.1 restores the complete reproducibility bundle, including analysis scripts, software-environment records, dependency specifications, and verification files omitted from the initial archive package. Scientific results, assignments, predictions, and conclusions are unchanged.