摘要：SummaryHuman leukocyte antigen (HLA) presentation of peptides is a prerequisite of T cell immune activation. The understanding of the rules defining this event has large implications for our knowledge of basic immunology and for the rational design of immuno-therapeutics and vaccines. Historically, most of the available prediction methods have been solely focused on the information related to antigen processing and presentation. Recent work has, however, demonstrated that method performance can be boosted by integrating information related to antigen abundance. Here we expand on these later findings and develop an extended version of NetMHCpan, called NetMHCpanExp, integrating information on antigen abundance from RNA-Seq experiments. In line with earlier works, the model demonstrates improved performance for both HLA ligand and cancer neoantigen epitope prediction. Optimal results are obtained by use of sample-specific abundance information but also reference datasets can be applied with a limited performance drop. The developed tool is available athttps://services.healthtech.dtu.dk/service.php?NetMHCpanExp-1.0.Graphical abstractDisplay OmittedHighlights•NetMHCpanExp, an extension of NetMHCpan, integrates antigen abundance data•NetMHCpanExp is built upon a modified version of NNAlign_MA•Minor performance loss when using reference instead of sample-specific RNA-Seq assays•Suboptimal MHC-I binders are “rescued” if arising from highly expressed proteinsImmunology; Biocomputational method; Transcriptomics; Biological sciences tools
