MHCII-peptide presentation: an assessment of the state-of-the-art prediction methods

Abstract

Major histocompatibility complex Class II (MHCII) proteins initiate and regulate immune responses by presentation of antigenic peptides to CD4+ T-cells and self-restriction. The interactions between MHCII and peptides determine the specificity of the immune response and are crucial in immunotherapy and cancer vaccine design. With the ever-increasing amount of MHCII-peptide binding data available, many computational approaches have been developed for MHCII-peptide interaction prediction over the last decade. There is thus an urgent need to provide an up-to-date overview and assessment of these newly developed computational methods. To benchmark the prediction performance of these methods, we constructed an independent dataset containing binding and non-binding peptides to 20 human MHCII protein allotypes from the Immune Epitope Database, covering DP, DR and DQ alleles. After collecting 11 known predictors up to January 2022, we evaluated those available through a webserver or standalone packages on this independent dataset. The benchmarking results show that MixMHC2pred and NetMHCIIpan-4.1 achieve the best performance among all predictors. In general, newly developed methods perform better than older ones due to the rapid expansion of data on which they are trained and the development of deep learning algorithms. Our manuscript not only draws a full picture of the state-of-art of MHCII-peptide binding prediction, but also guides researchers in the choice among the different predictors. More importantly, it will inspire biomedical researchers in both academia and industry for the future developments in this field.

Keywords: MHCII; bioinformatics; immunology; machine learning; peptide binding prediction; webserver.

Figure 1

Overall workflow of the construction of the benchmark dataset: we first collected all peptides from IEDB and selected data after 2020 (2020-2022) as our starting dataset Dset_ini . We removed the overlap between Dset_ini and the IEDB dataset entries before 2020 to make the positive Dset_pos dataset containing 67061 binding peptides of 11 to 19 amino acids; they span 20 allotypes. We used the non-bound regions of the host proteins to generate non-binding peptides of the same length as the binding ones. After that, we removed the non-binding peptides that appear in IEDB. The remaining peptides constitute the negative Dset_neg dataset of 67163 non-binding peptides. The union of Dset_pos and Dset_neg is the final benchmark dataset Dset_bench .

Figure 2

Overview of the computational approaches for MHCII-peptide interaction prediction. There are three types of predictors: scoring functions, machine-learning based tools, and consensus approaches. For each type of method, there are generally five steps to build a reliable predictor: data acquisition and preprocessing, feature generation and selection, model construction and optimization, performance evaluation, and the construction of a web server or independent software.

Figure 3

Analysis of the amino acid positional preferences for allotype-specific peptide ligands in specific HLAII allotypes: (A) HLA-DPA1*01:03 - DPB1*04:01,(B) HLA-DPA1*02:01-DPB1*14:01, (C) HLA-DPA1*01:03-DPB1*02:01, (D) HLA-DPA1*01:03-DPB1*03:01, (E) HLA-DPA1*01:03-DPB1*04:02, (F) HLA-DPA1*02:02-DPB1*05:01. The motifs are aligned with respect to the center of the 9-mer binding core. The overall height of letters indicates the sequence conservation at that position, while the height of each amino acid represents the relative frequency of that residue.

Figure 4

Performance of eleven MHC class II prediction methods assessed by their ROC curves and AUC values. The curves were generated by plotting the true positive rate (y-axis) against the false positive rate (x-axis). ROC curves for peptides binding to HLAII molecules specific for (A) HLA-DPA1*01:03-DPB1*02:01 (15mer), (B) HLA-DPA1*01:03-DPB1*04:01 (15mer) and (C) HLA-DPA1*02:01-DPB1*14:01 (15mer). AUC values for each prediction method are provided between parentheses in the subfigure legends. For prediction results on other HLA classes and peptide lengths see Supplementary Figure S1 and Data S4 .

Figure 5

Performance of MHC class II prediction methods for HLA-DR genes assessed by their ROC curves and AUC values. ROC curves for peptides binding to HLAII molecules specific for (A) HLA-DRB1*03:01 (15mer), (B) HLA-DRB1*04:01 (15mer), (C) HLA-DRB1*15:01(15mer), (D) HLA-DRB5*01:01 (15mer). AUROC values for each prediction method are provided between parentheses in the subfigure legends. For prediction results on other HLA classes and peptide lengths see Supplementary Figure S1 and Data S4 .