{"architecture":{"backbone":"64-unit LSTM, one network per allele","family":"LSTM recurrent network","input":"peptide sequence of any length (no alignment)","notes_en":"Transfer learning in two axes: from HLA-A*02:01 to every other allele, and from affinity data to mass-spectrometry data.","output":"predicted binding / presentation score"},"article":"/blog/modele-prezentacji-antygenu","cancer_slugs":["pan-cancer"],"category":"immunopeptidomics","confidence":"medium","datasets":[{"name":"IEDB \u2014 Immune Epitope Database","note":"","role":"training","slug":"iedb"},{"name":"IEDB automated benchmark (MHC class I)","note":"","role":"benchmark","slug":"iedb-benchmark"}],"developer":"Karchin lab, Johns Hopkins University","evaluation":[{"benchmark":"top 0.1% of a 1:999 decoy list","external":false,"metric":"PPV","source":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7056596/","value":"0.42 as reported by the authors \u2014 computed differently from NetMHCpan's 0.83, and not the same quantity"}],"hf":null,"kind":"task-model","license":"apache-2.0","limitations":"- Not pan-specific: an unseen allele is served by a neighbour's network, chosen by supertype.\n- Per-allele networks inherit the extreme data imbalance directly.","links":{"demo":null,"docs":null,"doi":"10.1158/2326-6066.CIR-19-0464","github":"https://github.com/KarchinLab/mhcnuggets","huggingface":null,"paper":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7056596/","pmid":null},"modalities":["protein-sequence","immunopeptidomics"],"name":"MHCnuggets","notable_uses":"","openness":"open-weights","regulatory":{"intended_use_en":"Research software.","intended_use_pl":"Oprogramowanie badawcze.","source_url":"https://github.com/KarchinLab/mhcnuggets","status":"research-only"},"regulatory_status":"research-only","release_date":"2020-03-01","run_snippet":"pip install mhcnuggets\nfrom mhcnuggets.src.predict import predict\npredict(class_='I', peptides_path='peptides.peps', mhc='HLA-A02:01')","settings":["basic-research","vaccine-design","immunotherapy"],"slug":"mhcnuggets","sources":[{"label":"Shao XM et al. High-Throughput Prediction of MHC Class I and II Neoantigens with MHCnuggets. Cancer Immunol Res 2020","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7056596/"},{"label":"cancer3.ai \u2014 Trzydzie\u015bci cztery litery zamka","url":"https://cancer3.ai/blog/modele-prezentacji-antygenu"}],"summary":"Allele-specific LSTM networks \u2014 one per allele, 148 for class I \u2014 that read peptides letter by letter, so no alignment or padding is needed, and that are trained by transfer learning from the data-richest allele.","tasks":["peptide-mhc-binding","antigen-presentation","neoantigen-prioritisation"],"training":{"summary_en":"IEDB binding affinities followed by fine-tuning on MS-eluted ligands; 148 class I and additional class II networks.","summary_pl":"Powinowactwa z IEDB, potem dostrajanie na ligandach ze spektrometru; 148 sieci klasy I i dodatkowe dla klasy II."},"updated_at":"2026-09-05T22:26:00.910427","url":"/ai-oncology/models/mhcnuggets","usage":{"library":"Python (pip install mhcnuggets)","snippet":"pip install mhcnuggets\nfrom mhcnuggets.src.predict import predict\npredict(class_='I', peptides_path='peptides.peps', mhc='HLA-A02:01')"},"verified_at":"2026-09-05T22:26:00.909264","verified_by":"editorial","version":"2.x","what_it_does":"Handles class I and class II in one framework and scales to high-throughput neoantigen calling; for an allele with no data it starts from the network of the closest allele by supertype."}