Computational Antigen Optimization through Symbolic Optimization and Affinity Maturation Simulation

Jonathan G. Faris, Mikel Landajuela, Kayla G. Sprenger, Daniel faissol, Felipe Leno da Silva

Keywords: symbolic optimization, antigen design, vaccine design, HIV, affinity maturation

TL;DR: Herein, we present a novel antigen optimization pipeline which utilizes symbolic optimization and affinity maturation simulations to generate and evaluate novel HIV antigen sequences.

Abstract: With the recent, significant improvement of computational tools for protein interaction prediction, the use of machine learning to support the development of vaccination regimens brings with it new hope for diseases which, so far, have eluded our best efforts at finding a cure, like HIV. We here propose BIOVAX, a novel pipeline combining symbolic optimization with affinity maturation simulation to generate highly-optimized antigens intended for vaccination development. We perform an in silico evaluation using real HIV targets, and show that the antigen designed by BIOVAX elicit estimated antibodies that bind more strongly to a diverse, global panel of real HIV viruses than both the parent sequence, and other computationally-designed antigen baselines available in the literature. BIOVAX is our first step towards a new generation of AI-assisted vaccine development pipelines.

Submission Number: 21