Prediction of T-cell epitopes

CTL epitopes using the NetCTL 1.2 server⁸⁴ (https://www.cbs.dtu.dk/services/NetCTL/) and the Immune Epitope Database (IEDB) servers⁸⁵ (https://tools.iedb.org/mhci/) were predicted. Epitope prediction by NetCTL is performed with 94–99% specificity and 54–89% sensitivity⁸⁶. In this study, the most prevalent HLA class I alleles (A1, A2, A3, A24, A26, B7, B8, B27, B39, B44, B58, and B62) were tested on all nine amino acid peptide sequences to predict and evaluate potential CTL epitopes⁸⁷. For the TAP transporter, epitope recognition, and C-terminal proteasomal cleavage parameters, the default thresholds of 0.15, 0.05, and 0.75 were used, respectively³⁶. Additionally, another set of epitopes for all HLA class I alleles in the IEDB was identified by the stabilized matrix method (SMM). Epitopes with percentile rank ≤ 2, IC₅₀ < 200 nM, and high rank were investigated and considered strong binders⁸⁸. Epitopes predicted for different alleles by NetCTL 1.2 and IEDB servers were selected for further analysis. Based on both servers' outcomes, common epitopes predicted by multiple alleles and favorable for the desired indices were picked for additional analysis.

On the other hand, HTL epitopes with a length of 15 residues were predicted by HLA class II alleles, including human HLA-DR, HLA-DP, and HLA-DQ alleles, using NetMHC II pan 3.2 server⁸⁹. Based on the determined percentiles of 2, 10, and more than 10%, the presented peptides were categorized as strong, moderate, and non-adhesive, respectively³⁶. Using the SMM-align method (NetMHCII 1.1) in the server IEDB⁹⁰, another group of HTL epitopes of the same length by 54 HLA class II alleles from the set of HLA-DR, HLA-DQ, and HLA-DP alleles were identified⁸⁵. Epitopes with IC₅₀ < 200 nM, percentile rank ≤ 2, and predicted by multiple alleles by both methods as strong binders were considered for further analyses. IEDB predicts the peptide binding to each MHC-II molecule using artificial neural networks (ANN) on a dataset trained with more than 500,000 binding affinity (BA) and eluted ligand mass spectrometry (EL) measurements and provides reliable results⁸⁸. Also, using the IFNepitope⁹¹ (https://crdd.osdd.net/raghava/ifnepitope/), IL4pred⁹² (https://webs.iiitd.edu.in/raghava/il4pred/design.php) and IL10pred⁹³ (https://webs.iiitd.edu.in/raghava/il10pred/predict3.php), HTL epitopes that specifically induce IFN-γ, IL-4 and IL-10 have been evaluated. The IFNepitope server detects with a maximum accuracy of 81.39% of all MHC class II binding overlapping peptides in a protein or antigen capable of inducing IFN-γ from CD4⁺ T-cells through methods such as machine learning strategy, motif-based analysis, and hybrid integrity⁹⁴.

For further analysis, each selected epitope (CTL and HTL) was subjected to a series of selectivity filters (i.e., immunogenicity, antigenicity, allergenicity, and toxicity), which were predicted with the aid of an IEDB class I immunogenicity server, VaxiJen v2.0, AllerTOP 2.0 and ToxinPred, respectively.