BCell-Epitope-Prediction: A local, reproducible pipeline for multi-epitope vaccine design: integration and structural validation against a panel of PDB antigen-antibody complexes

Abstract

Epitope-based subunit vaccine design requires chaining together heterogeneous analyses: B-cell antigenicity, topological accessibility, self-tolerance, allergenicity, T-helper/T-cytotoxic immunogenicity, and literature corroboration, historically scattered across separate, often web-only tools. This work presents a computational pipeline that integrates fourteen external tools across eight main phases, runs locally without network calls, and uses content-based self-caching. B-cell antigenicity is assessed via four engines (BepiPred-3.0, EpiDope, DiscoTope-3.0, and ScanNet), whose per-residue scores are fused into a single signal before thresholding. Validation used a panel of seventeen real antigen-antibody complexes from the Protein Data Bank spanning five pathogenic species, with ground truth from structural contact and independent literature verification. Fusing the four engines' scores, completed to a fixed candidate count, outperforms the best individual engine in F1 while reducing predicted residue volume. T-helper and T-cytotoxic immunogenicity are evaluated over each antigen's complete sequence, independently of the B-cell candidates. The final construct retains a B-cell epitope in 13 of 17 structures, and a T-cell block in 13 of 17 antigens, of which 6 match literature-documented epitopes and 2 more show partial correspondence.