Computational Inference Methods for HIV-1 Selective Sweeps Shaped by Early Cytotoxic T-Lymphocyte Response

Cytotoxic T lymphocytes (CTLs) play an important role in shaping HIV-1 infection. In particular, during the first weeks of infection, CTLs select for multiple escape mutations in the infecting HIV population. In recent years, methods have been developed that use intra-patient escape mutation data to estimate rates of escape from CTL selection. The resultant escape rate estimates have been used to study CTL kill rates and fitness costs associated with particular mutations, thereby providing a quantitative framework for exploring CTL response and HIV dynamics using patient datasets. Current methods for escape rate inference focus on a specific HIV mutant selected by a single CTL response. However, recent studies have shown that during the first weeks of infection, CTL responses occur at 1 − 3 epitopes and HIV escape occurs through complex mutation pathways. In this work we develop a model of initial infection, based on the well-known standard model, that allows us to model the complex mutation pathways of HIV escape. Under this model, we develop two computational inference methods. In one, we use a Bayesian approach to construct posteriors for the parameters of our model. In the second, we develop methods for hypothesis testing under our model. The methods are applied to two CHAVI datasets, demonstrating the importance of taking into account the interaction of multiple mutant variants and multi-directional selection.