MHCquant2 refines immunopeptidomics tumor antigen discovery

Abstract

Confident identification of human leukocyte antigen (HLA)-presented peptides is crucial for advancing cancer immunotherapy. We present MHCquant2, a scalable and modular Nextflow pipeline integrated into nf-core as a reproducible, portable, and standardized workflow for immunopeptidomics. This integration allows a community-driven, robust solution for high-throughput analyses across operating systems and cloud infrastructures. MHCquant2 integrates open-source tools including OpenMS, DeepLC, and MS²PIP, improving peptide identifications by up to 27% across diverse MS platforms, particularly enriching low-abundant peptides. MHCquant2 demonstrates state-of-the-art performance on our novel benign_MHCquant2 dataset (n = 92) and expands the benign human immunopeptidome by over 160,000 unique naturally presented HLA peptides.

Keywords: HLA; Immunopeptidomics; Immunotherapy; Mass spectrometry; Nextflow; Nf-core; Open-source; Pipeline.

Fig. 1

MHCquant2 workflow and HLA class I benchmark. A Subway plot of MHCquant2 key components. Each stop indicates a mandatory (o) or optional (▫) module of the pipeline. The dark gray subway represents the core workflow and the light gray subway the path of the quantification workflow. Stops indicated in black were implemented or extensively reworked within MHCquant2. B-H Benchmark against MHCquant1 using the HLA Ligand Atlas. B Total number of unique HLA class I binders across all samples and (C) distribution per sample identified using MHCquant1 and MHCquant2. D Total number of unique HLA class I binders and (E) distribution per sample identified without and with the feature generators DeepLC [20], MS²PIP [21], and their combination. Boxplots indicate the first to third quartile. Whiskers are defined as 1.5*IQR from the first and third quartile. F UpSet plot of total unique HLA class I binders identified by MHCquant1 and MHCquant2 shown at the top and their respective peptide intensity distribution displayed as a violin plot below. The inner boxplot of the violin plot depicts the median, first to third quartile of the distribution. G Density plot of Percolator q-value used as the FDR metric for MHCquant1 and MHCquant2. H Unique source proteins of HLA class I binders per tissue. Abbreviations: n.a., not available; FDR, false-discovery rate; IQR, interquartile range; HLA, human leukocyte antigen

Fig. 2

HLA class I immunopeptidome benchmark of MHCquant2, FragPipe, and PEAKS. The benchmark dataset (benign_MHCquant2) was generated from various benign primary tissues. Metadata describing the cohort is documented in SDRF format (Additional file 2: Table S1) A Sample overview and HLA class I binder yield of benign_MHCquant2 dataset. B HLA class I allotype distribution of all samples (N = 92). C Boxplot showing the distribution of HLA class I predicted binders and measured peptides ratio (purity) per charge state in the benign_MHCquant2 dataset. The subplot above the boxplot indicates the number of HLA class I peptides per charge. D Total number and (E) allotype annotated unique HLA class I binders identified using FragPipe, PEAKS, and MHCquant2 with the benign_MHCquant2 dataset. F UpSet plot of identified HLA class I binders by FragPipe, PEAKS, and MHCquant2 (left). Cumulative density plot of NetMHCpan percentile ranks for pipeline-exclusive peptides with indicated SB and WB threshold (middle). Length distribution of pipeline-exclusive (bar) and total (line) HLA class I binders (right). Abbreviations: SDRF, Sample and Data Relationship Format; HLA, human leukocyte antigen; SB, strong binder; WB, weak binder

Fig. 3

Refined tumor antigen discovery using MHCquant2. A Venn diagram depicting the HLA class I binder and HLA class II peptide overlap between the HLA Ligand Atlas [35], the Hoenisch Gravel et al. (PXD038782) dataset [18], and the benign_MHCquant2 dataset. B Stacked bar plots showing the contribution of benign_MHCquant2 HLA class I binders to public datasets according to primary tissue origin. C Comparison of published TAAs of AML [11], CLL [12], and OvCa [38] with re-analyzed TAAs by MHCquant2 and TAAs now identified in the new benign dataset. TAAs were defined according to the published filter criteria. D Sample frequency of shared HLA class I TAAs proposed by previous studies and identified by MHCquant2 for AML, CLL, and OvCa. TAAs are ranked according to sample frequency. E Intensity distribution of MHCquant2-identified peptides and neoepitopes of the melanoma dataset [8] compared to MHCquant1-identified neoepitopes in the respective dataset. MHCquant2-exclusive neoepitopes are annotated with their respective mutation location. F NetMHCpan percentile rank distribution of peptides and neoepitopes of the melanoma [8] dataset. G Mass-spectrometric neoantigen validation shown as mirror plot of experimentally eluted and synthetically validated spectrum of DVFPEGTRVGL (ENST00000353917 S296F, ENST00000360607 S337F, ENST00000372754 S419F, ENST00000372756 S378F) from one of the six detected neoepitopes in the melanoma dataset. Abbreviations: TAA, tumor-associated antigen; AML, acute myeloid leukemia; CLL, chronic lymphatic leukemia; OvCa, ovarian carcinoma; Mel, melanoma; HLA, human leukocyte antigen