Item

Abstract

The acquired immunodeficiency syndrome (AIDS) is one of the biggest medical challenges in the world today. Its causative pathogen, the human immunodeficiency virus (HIV), is responsible for millions of deaths per year. Although about two dozen antiviral drugs are currently available, progression of the disease can only be delayed but patients cannot be cured. In recent years, the new class of coreceptor antagonists has been added to the arsenal of antiretroviral drugs. These drugs block viral cell-entry by binding to one of the receptors the virus requires for infection of a cell. However, some HIV variants can also use another coreceptor so that coreceptor usage has to be tested before administration of the drug. This thesis analyzes the use of statistical learning methods to infer HIV coreceptor usage from viral genotype. Improvements over existing methods are achieved by using sequence information of so far not used genomic regions, next generation sequencing technologies, and by combining different existing prediction systems. In addition, HIV coreceptor usage prediction is analyzed with respect to clinical outcome in patients treated with coreceptor antagonists. The results demonstrate that inferring HIV coreceptor usage from viral genotype can be reliably used in daily routine.