HLAResolver: A statistical framework for resolving HLA ambiguity in T-cell response data

HLAResolver integrates MIRA response patterns with NetMHCpan predictions to resolve peptide–HLA ambiguity, achieving higher sensitivity at matched specificity while selecting fewer candidate HLAs per peptide.

• Why it matters: Training large-scale T-cell recognition models needs complete TCR–peptide–HLA assignments, but MIRA identifies TCRs without specifying which HLA presents each peptide, leaving multiple candidate HLAs.
• What they did: HLAResolver uses negative-binomial regression across HLA-typed donors to estimate allele–response associations, guided by NetMHCpan eluted-ligand predictions and sparse regularization, and is benchmarked on held-out MIRA experiments.
• The result: Compared with NetMHCpan and RATE, HLAResolver predicts observed MIRA responses with matched specificity, improves sensitivity, and assigns fewer candidate HLAs per peptide.

AI summary of the abstract · may contain errors · not medical advice · read the paper.