A central challenge in bioinformatics is the development of genomic scale methods for predicting protein structure, function and interactions. The difficulty of the general protein structure prediction problem suggests that one promising approach is to identify important subproblems that lend themselves to effective solutions. I will present two simplifications that my group has been pursuing in predicting protein interactions and protein structure.
First, I will discuss methods my group has developed for predicting protein-protein interactions mediated by the coiled-coil motif, an important motif that is found in proteins that participate in transcription, oncogenesis, cell structure, and cell-cell and viral-cell fusion events. We have introduced an optimization framework for predicting these types of protein interactions that uses both genomic sequence data and experimental data. Cross-validation tests show that the method is able to predict many aspects of protein-protein interactions mediated by the coiled-coil motif, and suggest that this methodology can be used as the basis for genome-wide prediction of coiled-coil protein interactions.
Second, I will discuss recent complexity results and computational methods for the side-chain positioning problem: given a fixed backbone and a protein sequence, predict the best conformation of the sequence's amino acids on this backbone. This is a widely studied problem with applications in homology modeling and protein design.
Created: 5/6/02 DAR