Developing A Novel Computational Framework To Predict Protein Interactions, Hot-Spots And Binding Affinities

Protein interactions play an important role in various biological processes, such as cell signaling, drug delivery and treating diseases. Predicting these interactions requires a fundamental understanding of protein-water interactions because every protein binding process involves disrupting the protein-water interactions, and replacing them by direct interactions between the binding proteins. Characterizing the protein-water interactions accurately, however, has proved to be challenging, because these interactions depend on not only the nanoscale topography, but also the precise chemical pattern presented by the protein surface. My thesis work aims to develop a novel method for characterizing the strength of protein- water interactions and predicting the interface through which proteins interact with one another. I built up the understanding of the fundamentals by investigating the interactions between water and simple model surfaces, such as hydrophobic carbon plates, self-assembled monolayers (SAM). In particular, I utilized the water density fluctuation based methods to characterize the free energetics of water near various surfaces, and quantitatively evaluated the hydrophobicity of surfaces. In addition, I study the context dependence of the surface hydrophobicity, showing that the local chemical patterns presented by a surface have a significant effect on the surface hydrophobicity, even when the chemistry content is kept the same. These fundamental studies inform and lead to the development of a novel framework to identify the potential interaction interfaces on proteins. The framework involves applying an unfavorable biasing potential to the entire protein hydration shell and locating the most hydrophobic spot on protein surfaces. These regions are the most probable interfaces through which proteins bind to other proteins or non-polar ligands, which then can be used to determine binding free energies. Finally, the proposed approach also enables determination of protein residues that contribute the most to the binding affinity, the so-called hot-spots. Since measuring the free energetics of the water density fluctuations requires the expen- sive non-Boltzmann sampling techniques such as umbrella sampling, it limits the size and amount of the proteins I can readily investigate. In order to overcome this computational burden, I also developed the sparse sampling method to efficiently compute the free energetics of water density fluctuations, making it roughly 2 orders of magnitude more efficient than umbrella sampling. The method enables me to characterize water density fluctuations in the entire hydration shell of the protein, ubiquitin, a large volume containing an average of more than 600 waters, as well as enormous spherical volumes in bulk water containing up to 30,000 waters on average. The sparse sampling method thus opens up the potential studies on larger and more complex protein systems