Randomized Algorithms, Bayesian Phylogenetics, and Computational Vaccine Design

A significant challenge in vaccine design is the elicitation of broadly-neutralizing antibodies (bnAbs) to protect against viral escape. We are developing computational models of affinity maturation to understand bottlenecks in the inducibility of bnAbs and guide sequential vaccine strategies. Doing so requires accounting for context-dependence in somatic hypermutation, which in turn requires phylogenetic inference in the presence of site-dependence, a long-standing challenge in computational biology. We describe an approximation algorithm for this problem using a combined data-augmentation and importance sampling scheme. We apply this approach to reconstruct maturation lineages from high-throughput B cell receptor repertoire sequencing data, and examine the impact of context-dependence on reconstruction accuracy. We then apply our models to the problem of choosing targets for the design of boosting immunogens aimed at elicitation of HIV bnAbs.