High-dimensional HIV-1 quasispecies modeling guides escape-proof antibody design

Abstract

Rapidly evolving viruses form diverse quasispecies that enable escape from immune responses and treatments. For example, HIV-1 can rebound within weeks of broadly neutralizing antibody (bNAb) treatment through the outgrowth of high-fitness escape mutants in the quasispecies or the evolution of new escape variants. Most existing models of viral dynamics consider only a small number of viral variants and either assume arbitrary mutant fitness distributions or require extensive fitting to sparse clinical data. Here, we develop a high-dimensional HIV-1 quasispecies model that captures the dynamics of millions of viral strains and parameterize this using in silico binding affinity predictions. Without fitting to experimental data, the model qualitatively reproduces viral rebound following bNAb treatment. Lower-dimensional model projections recover these dynamics only when informed by features derived from the high-dimensional model. Finally, we use the model to develop a quasispecies-based framework for antibody optimization and identify antibodies predicted to effectively suppress viremia. Together, our results demonstrate that integrating mechanistic genotype-phenotype maps with high-dimensional quasispecies models provides unprecedented insights into viral evolution.