Abstract: A key task in data mining is network reconstruction, i.e., given observed data from a system of interest, to identify the constraints underlying the system variables, and to summarize these constraints into a network model of the system. By drawing examples from protein design, neuroscience, and systems biology, we will demonstrate the central role of network reconstruction in data-driven computational science. Although the nature of data, the application goals, and the type of network to be mined are slightly different, we will show how techniques for learning graphical models can be adapted to address network reconstruction tasks in all of the above domains. Case studies are illustrated through specific problem families such as G-protein coupled receptors and WW domains (for protein design), multi-channel measurements from cortical neurons (for neuroscience), and the budding yeast cell cycle (for systems biology).

(This talk does not assume familiarity with the application domains and will be accessible to a computer science audience.)

Short Bio: Naren Ramakrishnan is an associate professor of computer science and faculty fellow in engineering at Virginia Tech. He also serves as an adjunct professor at the Institute of Bioinformatics and Applied Biotechnology (IBAB), Bangalore, India. His research interests include problem solving environments, mining scientific data, and information personalization. He is an area editor for IEEE Computer and was program chair for the Seventh IEEE International Conference on Data Mining (ICDM'07). He was recently named to Computerworld's list of '40 under 40' innovative IT people to watch (July 2007). Ramakrishnan received his Ph.D. in computer sciences from Purdue University.