Automated benchmarking of peptide-MHC class I binding predictions

Trolle, Thomas; Metushi, Imir G.; Greenbaum, Jason A.; Kim, Young S.; Sidney, John; Lund, Ole; Sette, Alessandro; Peters, Bjoern; Nielsen, Morten

doi:10.1093/bioinformatics/btv123

Cited by 134 publications

(150 citation statements)

References 36 publications

Supporting

Mentioning

148

Contrasting

Order By: Relevance

“…35,36,66 NetMHCpan was selected because it consistently performs as one of the best prediction tools across a wide array of alleles, and also provides predicted IC50 nM values for the complete set of common class I alleles considered here. 67,68 In addition to predicted affinity (IC50), NetMHCpan also provides a percentile score expressing the relative capacity of each peptide to bind each specific allele, compared to a universe of potential sequences of the same size. While newer versions of NetMHCpan are available, we focused on the 2.8 version as only for this implementation a length-rescaling model is available, as described by Trolle et al 40 Briefly, ligands from five common HLA class I alleles were eluted and the allele-specific length-distribution of ligands was determined.…”

Section: Methodsmentioning

confidence: 99%

Predicting T cell recognition of MHC class I restricted neoepitopes

et al. 2018

Self Cite

View full text Add to dashboard Cite

Epitopes that arise from a somatic mutation, also called neoepitopes, are now known to play a key role in cancer immunology and immunotherapy. Recent advances in high-throughput sequencing have made it possible to identify all mutations and thereby all potential neoepitope candidates in an individual cancer. However, most of these neoepitope candidates are not recognized by T cells of cancer patients when tested in vivo or in vitro, meaning they are not immunogenic. Especially in patients with a high mutational load, usually hundreds of potential neoepitopes are detected, highlighting the need to further narrow down this candidate list. In our study, we assembled a dataset of known, naturally processed, immunogenic neoepitopes to dissect the properties that make these neoepitopes immunogenic. The tools to use and thresholds to apply for prioritizing neoepitopes have so far been largely based on experience with epitope identification in other settings such as infectious disease and allergy. Here, we performed a detailed analysis on our dataset of curated immunogenic neoepitopes to establish the appropriate tools and thresholds in the cancer setting. To this end, we evaluated different predictors for parameters that play a role in a neoepitope’s immunogenicity and suggest that using binding predictions and length-rescaling yields the best performance in discriminating immunogenic neoepitopes from a background set of mutated peptides. We furthermore show that almost all neoepitopes had strong predicted binding affinities (as expected), but more surprisingly, the corresponding non-mutated peptides had nearly as high affinities. Our results provide a rational basis for parameters in neoepitope filtering approaches that are being commonly used.Abbreviations: SNV: single nucleotide variant; nsSNV: nonsynonymous single nucleotide variant; ROC: receiver operating characteristic; AUC: area under ROC curve; HLA: human leukocyte antigen; MHC: major histocompatibility complex; PD-1: Programmed cell death protein 1; PD-L1 or CTLA-4: cytotoxic T-lymphocyte associated protein 4

show abstract

Section: Methodsmentioning

confidence: 99%

Predicting T cell recognition of MHC class I restricted neoepitopes

et al. 2018

Self Cite

View full text Add to dashboard Cite

show abstract

“…For a given HLA molecule and a given peptide length, several benchmarks have shown that binding predictions correlate well with measured binding affinities (1–4), and that peptides with high predicted affinity contain the vast majority of T cell epitopes (5, 6). This has allowed comprehensive mapping of epitopes in entire pathogens by focusing testing on a manageable number of top predicted binders, saving vast amounts of resources (7–12).…”

Section: Introductionmentioning

confidence: 99%

The Length Distribution of Class I–Restricted T Cell Epitopes Is Determined by Both Peptide Supply and MHC Allele–Specific Binding Preference

Trolle

McMurtrey

Sidney

et al. 2016

The Journal of Immunology

Self Cite

190

196

View full text Add to dashboard Cite

HLA class I binding predictions are widely used to identify candidate peptide targets of human CD8+ T cell responses. Many such approaches focus exclusively on a limited range of peptide lengths, typically 9 and sometimes 9-10 amino acids, despite multiple examples of dominant epitopes of other lengths. Here, we examined if epitope predictions can be improved by incorporating the natural length distribution of HLA class I ligands. We found that while different HLA alleles have diverse length binding preferences, the length profiles of ligands that are naturally presented by these alleles are much more homogeneous. We hypothesized that this is due to a defined length profile of peptides available for HLA binding in the endoplasmic reticulum. Based on this, we created a model of HLA allele specific ligand length profiles, and demonstrate how this model, in combination with HLA binding predictions, greatly improves comprehensive identification of CD8+ T cell epitopes.

show abstract

“…To summarize briefly, a neural network is trained to output the affinity of a given pMHC pair that has the MHC represented by pseudo sequences constructed from HLA residues in contact with bound peptides, which are poly morphic in known functional HLA-A, HLA-B or HLA-C alleles. Pan-specific tools, such as NetMHCpan 77 and NetMHCIIpan 78 , scored among the best performers, even compared to allele-specific approaches 79,80 . However, although several assessments and comparisons have been made in the past 79,81,82 , there are currently no recent independent benchmark studies that can be used to recommend specific tools.…”

Section: Phased Genotypesmentioning

confidence: 99%

Computational genomics tools for dissecting tumour–immune cell interactions

et al. 2016

View full text Add to dashboard Cite

Recent breakthroughs in cancer immunotherapy and decreasing costs of high-throughput technologies have sparked intensive research into tumour-immune cell interactions using genomic tools. The wealth of the generated data and the added complexity pose considerable challenges and require computational tools to process, to analyse and to visualize the data. Recently, various tools have been developed and used to mine tumour immunologic and genomic data effectively and to provide novel mechanistic insights. Here, we review computational genomics tools for cancer immunology and provide information on the requirements and functionality in order to assist in the selection of tools and assembly of analytical pipelines.

show abstract

Automated benchmarking of peptide-MHC class I binding predictions

Abstract: Supplementary data are available at Bioinformatics online.

Cited by 134 publications

References 36 publications

Predicting T cell recognition of MHC class I restricted neoepitopes

Predicting T cell recognition of MHC class I restricted neoepitopes

The Length Distribution of Class I–Restricted T Cell Epitopes Is Determined by Both Peptide Supply and MHC Allele–Specific Binding Preference

Computational genomics tools for dissecting tumour–immune cell interactions

Contact Info

Product

Resources

About