Advances in linear epitope-based subunit vaccines powered by artificial intelligence: current status and challenges

Vollständiger Abstract

Worum geht es in dieser Arbeit?

In this review, linear epitope-based subunit vaccines are systematically profiled to outline their current developmental status, integrating recent advances in computational immunogenic epitope screening, site-specific chemical modification for stability enhancement, next-generation delivery platform matching, and ongoing clinical translation efforts. Deep learning-based MHC-binding prediction tools including NetMHCpan, MHCflurry 2.0, and MARIA, supported by robust empirical evidence, are rigorously evaluated, with notable improvements observed in the efficiency of identifying B/T cell linear epitopes with high immunogenicity, while it is also noted that the field still faces core bottlenecks such as weak in vivo immunogenicity, rapid enzymatic degradation, and insufficient cross-protection against viral variants. According to this integrative evidence-based assessment, feasible optimization pathways are further outlined, including the synergy between AI-driven epitope structural design and virus-like particle-based delivery systems, to provide practical references for the rational development of next-generation broad-spectrum linear epitope-based vaccines.