NetMHC

NetMHC refers to a family of computational tools used to predict binding of peptides to major histocompatibility complex (MHC) class I molecules. Originally based on artificial neural networks, NetMHC predicts the binding affinity of short peptides to specific MHC class I alleles, conveying results as predicted IC50 values and percentile ranks relative to a large peptide set. Over time, expanded versions such as NetMHCpan were developed to provide pan-allele predictions that cover thousands of MHC class I molecules across human and non-human species by leveraging sequence similarity.

NetMHC is available as a web server and as downloadable software for local installation. Users input a peptide sequence and select one or more MHC alleles; the output includes predicted binding affinity, percentile rank, and a binder classification (strong or weak) using commonly used thresholds, though definitions may vary by version. NetMHCpan extends the approach to include a broad range of HLA alleles by integrating binder data across known MHC sequences; NetMHCcons provides a consensus prediction by combining multiple predictors.

Typical applications include epitope mapping in vaccine design and identification of candidate neoantigens in cancer immunotherapy. Predictions favor peptides of nonamer length for class I binding, though lengths can be variable depending on allele.

Limitations include dependence on training data quality, allele representation biases, and the inherent uncertainties of in silico predictions. Experimental validation remains essential, and predictions should be interpreted alongside other factors, such as peptide processing and presentation.