Modeling HIV Escape from Neutralizing Antibody Respones

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): HIV can rapidly escape neutralizing antibody responses in vivo. Although escape facilitates viral persistence within the infected patient, escape from neutralization implies that the antibody response is exerting significant selective pressure on the virus. Consequently, a vaccine-boosted humoral response may be effective in preventing infection, as the small population of infecting viruses cannot evolve quickly. An effective vaccine will also be able to induce significant heterologous neutralization responses. Although HIV is generally resistant to neutralization by sera from HIV-positive patients, some individuals mount broadly reactive neutralization responses The recent availability of high-throughput assays to quantify neutralizing antibody responses coupled with the ability to sequence full-length clones of the HIV-1 envelope gene permits a quantitative analysis of autologous and heterologous neutralization responses, and their relationship to viral genetic variation. We propose to quantify the roles of point mutations and insertions/deletions in the HIV-1 envelope in viral escape based on data from individuals with primary HIV infection. Neutralizing antibody responses will be measured using a rapid, highly reproducible recombinant virus/reporter assay, and genetic variation in envelope will be measured using both population-based and clonal sequencing. Analysis of these data presents a number of statistical challenges. We will develop statistical models to (a) obtain a quantative description of the dynamics of autologous neutralizing antibody responses and the patterns of viral escape and (b) to quantify variation between patients' sera and between viruses in heterologous neutralization, and relate this variation to host and viral (e.g. viral genotype) factors. These models will take the hierarchical nature of HIV data into account, will use spline-based methods to capture nonlinear effects, and will incorporate submodels for censoring and measurement error.