Association mapping of complex phenotypes using models at a systems level

The goal of genetic association studies is to relate polymorphic genetic loci to quantitative traits that capture the phenotypic variation of interest. While this task is well studied for "simple" univariate phenotypes, recent high-dimensional phenotypes ask for more advanced modeling techniques at a systems level. In this talk I will discuss recent machine learning techniques to address these emerging challenges and illustrate them using two case studies. First, I will focus on QTL mapping of high-dimensional microarray data in yeast. In this setting, inference of unmeasured transcription factor activations from the expression levels and pathway information allows the phenotypic variation to be dissected at a previously unavailable level of detail. Second, I will discuss mapping techniques for the treatment response of human depressed patients. In these data, the joint modeling of observations that change over time and complex confounding influence need to be taken into account to identify relevant genetic markers