New tools for systems-level analysis of regulation and signaling dynamics
Speaker: Edoardo Airoldi , Harvard
Mapping the functional landscape driving complex cellular phenotypes is a central goal of modern genome- and proteome-scale studies. In this talk, I will present new statistical tools that support such analysis. First, we will consider perturbation experimental designs where a quantitative trait of interest Y, such as gene expression is measured multiple times, for different values of a covariate X, such as time or growth rate, and for different factors F, such as knock-out strains or growth media. In this context, I will introduce a linear model framework where the linear response of individual genes depends explicitly on the linear response of the functions these genes are annotated to, according to the gene ontology. We will use this model to study the functional basis of the cellular response to a series of nutrient perturbations, in yeast. Second, we will consider coordinated experimental designs where we observe multiple high-dimensional phenotypical responses Y1...Yk over time. In this context, we want to quantify the extent to which the newly generated data supports current hypotheses about regulation and signaling dynamics, and we want to generate novel hypotheses that can be tested at the bench. We will use this model to explore the pheromone response pathway, in yeast. If time allows, I will conclude with an overview of our new template-based technology to analyze high-throughput sequences that runs on Amazon EC2 and can be used to perform deep statistical analysis, genome-wide, in the order of minutes.