Abstract:
|
Vaccines to prevent or cure infection by pathogens have improved public health dramatically over the past two centuries. Determining the sufficient immune response for protection (or for reducing pathogen load) is an important part of vaccine development. I will present a statistical analysis of a study that constitutes the initial phase of the development of a therapeutic vaccine to prevent spread of acute tuburculosis (TB) infection in a non-human primate model. In this randomized controlled trial, rhesus cytomegalovirus (CMV) vectors modified to contain genes of the TB pathogen were delivered intravenously to TB negative rhesus macaques (RM), who were then infected with TB. The "correlates analysis" to detect associations between the extent of TB disease and the high-dimensional immune response data involved a number of characteristic challenges, including high correlation across predictors and shifting patterns of association over time and tissues. I will present a hybrid supervised learning procedure that I developed to address these challenges that incorporates both principal component analysis and significance testing in a frequentist procedure.
|