{"architecture":{"family":"neural network with peptide, expression and cleavage features","input":"peptide + allele + source-gene expression + flanking cleavage context","notes_en":"Adding expression is what a peptide-only model structurally cannot do; the price is that you need RNA data for the sample.","output":"presentation score / percentile"},"article":"/blog/modele-prezentacji-antygenu","cancer_slugs":["pan-cancer"],"category":"immunopeptidomics","confidence":"high","datasets":[{"name":"Mono-allelic HLA class I peptidome (Sarkizova / Abelin)","note":"this model and this dataset come from the same paper","role":"training","slug":"monoallelic-peptidome"}],"developer":"Broad Institute / Dana-Farber Cancer Institute (Keskin, Wu, Carr labs)","evaluation":[{"benchmark":"top 0.1% of a 1:999 decoy list","external":false,"metric":"PPV","source":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7008090/","value":"reported by the authors at 1:999 \u2014 a stricter setup than several competitors use"}],"hf":null,"kind":"task-model","license":"free web server; academic use","limitations":"- Web-only: not embeddable in an offline pipeline.\n- Expression features require matched RNA data, so it is not a universal 'peptide in, score out' tool.","links":{"demo":"http://hlathena.tools/","docs":null,"doi":"10.1038/s41587-019-0322-9","github":null,"huggingface":null,"paper":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7008090/","pmid":null},"modalities":["protein-sequence","immunopeptidomics","transcriptomics"],"name":"HLAthena","notable_uses":"The underlying allele panel covers at least one allele in 95% of people worldwide for each of HLA-A, -B and -C \u2014 the reason the 'long tail' of alleles shrank at all.","openness":"api-only","regulatory":{"intended_use_en":"Research use.","intended_use_pl":"Do bada\u0144.","source_url":"http://hlathena.tools/","status":"research-only"},"regulatory_status":"research-only","release_date":"2020-01-13","run_snippet":"","settings":["basic-research","vaccine-design","immunotherapy"],"slug":"hlathena","sources":[{"label":"Sarkizova S et al. A large peptidome dataset improves HLA class I epitope prediction. Nat Biotechnol 2020","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7008090/"},{"label":"cancer3.ai \u2014 Trzydzie\u015bci cztery litery zamka","url":"https://cancer3.ai/blog/modele-prezentacji-antygenu"}],"summary":"Presentation predictor trained on the mono-allelic peptidome of 95 cell lines \u2014 186,464 peptides across 95 HLA alleles, fifteen of which had no described motif before \u2014 which is the dataset that changed this field more than any architectural idea.","tasks":["antigen-presentation","peptide-mhc-binding","neoantigen-prioritisation"],"training":{"institutions":"Broad Institute, Dana-Farber","size":"186,464 peptides / 95 alleles","summary_en":"95 mono-allelic B721.221 lines (31 HLA-A, 40 HLA-B, 21 HLA-C, 3 HLA-G), 186,464 unique peptides, median 1,860 per allele.","summary_pl":"95 linii monoallelicznych B721.221 (31 HLA-A, 40 HLA-B, 21 HLA-C, 3 HLA-G), 186 464 unikalne peptydy, mediana 1860 na allel."},"updated_at":"2026-09-05T22:26:00.955691","url":"/ai-oncology/models/hlathena","usage":{"api":"http://hlathena.tools/","library":"web server","notes_en":"No standalone weights; use the web tool or reuse the published peptidome (see the data card) to train your own model."},"verified_at":"2026-09-05T22:26:00.948597","verified_by":"editorial","version":null,"what_it_does":"Scores peptide\u2013allele pairs and adds features beyond the peptide: gene expression of the source protein and cleavage context, which is why its predictions track real presentation more closely than binding-only tools."}