Development of a simulation algorithm and scoring functions for predicting biomolecular interactions

Abstract

Simulations based on the knowledge of various interactions working at the atomic level have been successfully applied to studying the structure and function of biomolecules. Energy functions (ranking functions) are at the basis of such simulations, providing the thermodynamic criteria for predicting the correct structure of proteins and explain why they assume such structures. This study involves the application of simulation and modeling to two critical problems and the development of new algorithms and ranking functions for improved results. Chapter 1 reports the development of a new method to predict the structure of peptide-MHC complexes, which form the basis of self/non-self discrimination. The immune system protects the body by selectively removing potentially harmful matter. Especially, proteins are under surveillance at the amino acid level for their various activities as the workhorses of life. This surveillance is realized the recognition by T cells of processed peptides that are bound and displayed on the MHC molecules. While the prediction of peptide-MHC complexation has been actively studied, an in-depth analysis including the TCR is not being realized due to the lack of means to obtain structural information of the numerous potential binding pairs. Therefore, an improved algorithm for simulation and new ranking functions were developed to predict the structure of peptide-MHC complexes. The new simulation samples candidates broader and faster, and the new ranking functions accurately selects native-like conformations. Chapter 2 describes the application of computational approaches to alter the substrate selectivity of a quorum quenching enzyme (QQE), which has the potential as an alternative to antibiotics. A broad range of microbes control numerous activities via chemical quorum signaling, which includes the expression of virulence. Therefore, enzymes which remove the signaling molecules can be used to prevent the expression of virulence without evolving resistance. However, natural QQEs exhibit broad substrate specificity, making target-specific control difficult. Also, those which degrade AHLs (N-homoserine lactones) do not efficiently remove short-chain AHLs. Therefore, the substrate preference of wild-type AiiA for long-chain AHLs was reversed by engineering. To this aim, molecular docking was performed to predict the modes of enzyme-substrate binding. Mutants were designed based on the predictions, and they displayed at most 100 times shifted substrate preference. In addition, molecular dynamics simulation was performed to reveal new clues in a dynamic context of enzyme-substrate interactions and a possibility of virtual screening of mutant enzymes using potential energy.