Alex Rodionov, Jonathan Rose, Elisabeth R.M. Tillier and Alexandr Bezginov
Department of Electrical and Computer Engineering, and Department of Medical Biophysics and Ontario Cancer Institute, University of Toronto
October, 2010
Proteins are the building blocks of all living organisms, and studying protein interactions helps us understand how life works. Lab experiments and computer simulations of protein interactions are expensive, so it is advantageous to predict ahead of time which protein pairs are likely to interact at all, before undertaking detailed experiments.
This talk will be about my current Masters thesis, which is an interdisciplinary effort involving biology, software, and FPGA hardware design. Our collaborators at the Department of Medical Biophysics have developed a new method called MatrixMatchMaker (MMM) for predicting protein interactions by looking for a phenomenon called co-evolution between pairs of proteins. I will describe MMM and our work in accelerating the original MMM algorithm using a combination of algorithmic and software optimization, and ongoing work in hardware acceleration.