Specific interactions among biomolecules are important for proper cellular function. Yet the prevalence of large, paralogous protein families poses a challenge to discrimination by molecular recognition: many potential interactions partners share a common evolutionary history and thus have similar sequences and structures. Using closely integrated computational and experimental methods, we have been investigating sequence and structural determinants of interaction specificity in evolutionarily conserved interaction domains. We use alpha-helical coiled coils, Bcl-2 family proteins, and two-component signaling histidine kinase dimers as models, with the goal of learning to predict and design specific complexes involving these components. I will discuss different approaches we have taken to modeling specificity. Purely computational structure-based models are generally not yet usefully predictive, though such approaches are important for protein design. Consistently, we find that experimental data are critical for achieving useful models, and my group is exploring ways that computation and experiments can be ever more effectively combined.
Copyright. All Rights Reserved.