Systems-scale and integrative "omic" analysis of host-pathogen interactions in malaria

Mark Styczynski (April 9, 2014)

Please install the Flash Plugin


As part of the Malaria Host-Pathogen Interaction Center, our goal is to study and model the response of both host and pathogen to the course of malarial infection, treatment, and recurrence or recrudescence, using multiple levels of "omic" data. Detailed mathematical models are a desired ultimate product of our study of malaria, and while there are certainly some intuitive candidate systems for such modeling, it is not necessarily clear a priori which other systems should be modeled, nor which variables are important to include in those models. Our goal is to exploit the multiple levels of systems-scale datasets being generated in our center to identify such candidates for detailed models and follow-up experiments.

The main task in achieving this goal is discovery of novel, unknown interactions between our measured variables. This can be accomplished via a number of classes of approaches, including statistical analyses and machine learning. Here, we will focus on our machine learning approaches to identifying subnetworks of interesting interactions, specifically using probabilistic graphical models to construct interaction networks. Within this domain, two of the biggest obstacles to accomplishing our goal are 1) computational tractability given the high dimensional variable space, and 2) integrating multiple disparate data types, each with potentially different scales of variable space dimensionality (tens of measurements vs. tens of thousands of measurements) and different time scales, such that no data type dominates or is dwarfed in importance. We will present our algorithmic work addressing these problems, along with applications to the malaria data that has been generated by our center to date.