Unsupervised Mining of HLA-I Peptidomes Reveals New Binding Motifs and Potential False Positives in the Community Database

Abstract

Modern vaccine designs and studies of human leukocyte antigen (HLA)-mediated immune responses rely heavily on the knowledge of HLA allele-specific binding motifs and computational prediction of HLA-peptide binding affinity. Breakthroughs in HLA peptidomics have considerably expanded the databases of natural HLA ligands and enabled detailed characterizations of HLA-peptide binding specificity. However, cautions must be made when analyzing HLA peptidomics data because identified peptides may be contaminants in mass spectrometry or may weakly bind to the HLA molecules. Here, a hybrid de novo peptide sequencing approach was applied to large-scale mono-allelic HLA peptidomics datasets to uncover new ligands and refine current knowledge of HLA binding motifs. Up to 12-40% of the peptidomics data were low-binding affinity peptides with an arginine or a lysine at the C-terminus and likely to be tryptic peptide contaminants. Thousands of these peptides have been reported in a community database as legitimate ligands and might be erroneously used for training prediction models. Furthermore, unsupervised clustering of identified ligands revealed additional binding motifs for several HLA class I alleles and effectively isolated outliers that were experimentally confirmed to be false positives. Overall, our findings expanded the knowledge of HLA binding specificity and advocated for more rigorous interpretation of HLA peptidomics data that will ensure the high validity of community HLA ligandome databases.

Improved predictions of antigen presentation and TCR recognition with MixMHCpred2.2 and PRIME2.0 reveal potent SARS-CoV-2 CD8+ T-cell epitopes

“…In addition to determining binding motifs and peptide length distributions for the different alleles expressed in a sample, motif deconvolution is useful to identify potential sources of noise in the data. 18 , 23 , 34 Noise in HLA-I peptidomics data can consist of peptides from the same sample (e.g., contaminants pulled down together with HLA-I ligands, but not binding to HLA-I molecules), peptides from other samples (e.g., contaminants due to suboptimal cleaning of MS equipment), or wrongly identified peptides occurring during the computational annotation of mass spectra.…”

Deep learning methods for de novo peptide sequencing

“…venomics [90,94,95], and metaproteomics [78,97]. A variety of other studies use de novo sequencing tools to detect various classes of short or unexpected peptide sequences [80,82,83,85,92,93,96,101,102]. We also note that all seven of the tools have been used at least once in a study published independently of the original authors, suggesting that the software can be successfully used by others.…”

MHCSeqNet2—improved peptide-class I MHC binding prediction for alleles with low data

“…The first HLA binding dataset comes from combining several mass spectrometry-based mono-allelic HLA peptidomics studies ( Abelin et al 2017 , 2019, Marco et al 2017 , Solleder et al 2019 , Sarkizova et al 2020 ) with peptide–HLA pairs curated by the Immune Epitope Database (IEDB) ( Vita et al 2018 ). Duplicated peptide–HLA pairs and peptides with modifications were removed.…”