This tool predicts the function of peptides: submit one or more peptide sequences and get a binary prediction for each of 22 functions in a single run (each with a probability, a decision threshold, and a positive or negative call). It is intended for the first-pass activity screening and safety assessment of peptide drugs.
The 22 functions are grouped into 6 categories by application area: Activity — Antimicrobial (AMP): antimicrobial, antibacterial, antifungal, antiparasitic, antiviral; Activity — Anticancer: anticancer, tumor T-cell antigen (ttca); Activity — Metabolic Regulation: ACE inhibition, DPP-IV inhibition, antidiabetic; Activity — Other: antiaging, anti-inflammatory, antioxidant, neuropeptide, quorum sensing; ADME / Drug-likeness: nonfouling, cell-penetrating (cpp), blood-brain barrier penetration (bbp); Safety (Tox): allergenicity, hemolytic activity, neurotoxicity, toxicity.
The model was trained on PepBenchData-50 with the official hybrid split (8:1:1, 5 seeds), and the decision thresholds were tuned on the validation set.
Scope and limitations: only linear peptides composed of the 20 standard amino acids are covered (non-natural peptides and SMILES input are not supported), and continuous-value endpoints such as MIC, HC50 and PAMPA are not predicted. A sequence length of ≤ 50 aa is recommended.
1. Peptide Sequences (up to 10 FASTA entries):
Parsed sequences: 0, total residues: 0
Model Performance Metrics
PepBench 22-task Summary (hybrid split, 5 seeds, mean±std)
============================================================================
ace_inhibitory MCC=0.4915±0.0383 AUC=0.8334±0.0162 THR=0.5300±0.2308
allergen MCC=0.6501±0.0497 AUC=0.9008±0.0163 THR=0.3300±0.1255
antiaging MCC=0.1631±0.1172 AUC=0.5890±0.0786 THR=0.7900±0.0894
antibacterial MCC=0.6019±0.0206 AUC=0.9232±0.0029 THR=0.6100±0.0548
anticancer MCC=0.6240±0.0173 AUC=0.9205±0.0067 THR=0.4900±0.0962
antidiabetic MCC=0.3467±0.0857 AUC=0.7741±0.0400 THR=0.3900±0.2074
antifungal MCC=0.5916±0.0420 AUC=0.9097±0.0111 THR=0.5100±0.0652
antiinflamatory MCC=0.4383±0.0528 AUC=0.7957±0.0226 THR=0.3900±0.1294
antimicrobial MCC=0.6176±0.0158 AUC=0.9201±0.0045 THR=0.6000±0.0354
antioxidant MCC=0.3439±0.0771 AUC=0.7218±0.0370 THR=0.6400±0.1917
antiparasitic MCC=0.5835±0.0510 AUC=0.9080±0.0144 THR=0.6800±0.1304
antiviral MCC=0.5594±0.0313 AUC=0.8835±0.0081 THR=0.5800±0.1255
bbp MCC=0.3653±0.0774 AUC=0.6885±0.0770 THR=0.7400±0.1140
cpp MCC=0.4910±0.1183 AUC=0.8562±0.0370 THR=0.6100±0.2275
dppiv_inhibitors MCC=0.5339±0.0351 AUC=0.8366±0.0212 THR=0.5400±0.1710
hemolytic MCC=0.5412±0.0293 AUC=0.8596±0.0158 THR=0.5600±0.2815
neuropeptide MCC=0.5505±0.0358 AUC=0.8705±0.0122 THR=0.5300±0.1304
neurotoxin MCC=0.3394±0.0829 AUC=0.7251±0.0412 THR=0.5900±0.1673
nonfouling MCC=0.3919±0.0469 AUC=0.7721±0.0172 THR=0.4100±0.0962
quorum_sensing MCC=0.5655±0.1136 AUC=0.8483±0.0405 THR=0.7000±0.1275
toxicity MCC=0.3688±0.0695 AUC=0.7486±0.0221 THR=0.4900±0.1387
ttca MCC=0.5169±0.1156 AUC=0.8591±0.0497 THR=0.5600±0.1673
----------------------------------------------------------------------------
macro_avg_MCC 0.4853 ± 0.0092
macro_avg_AUC 0.8247 ± 0.0067
macro_avg_THR 0.5577 ± 0.0213
============================================================================
Note: the hybrid split combines the k-mer split and the mmseq split; THR is the decision threshold tuned on the validation set (mean over 5 seeds).
References
Zhang J, Wang R, Zhou K, et al. PepBenchmark: a standardized benchmark for peptide machine learning. Presented at: International Conference on Learning Representations; 2026. https://openreview.net/forum?id=NskQgtSdll
Last updated: 2026-09-21