Deciphering inter- and intra-molecular interactions based on the dynamic and structural properties of molecules

Abstract

Inter- and intra-molecular interactions play key roles in regulating molecular processes within a cell. Decoding of these molecular interaction patterns can be guided by the important properties of molecules. DNA and protein molecules are dynamic objects and their conformational changes normally trigger functional changes. Thus, dynamic and structural properties of molecules are crucial elements in deciphering their molecular interactions. This thesis examines different types of molecular interactions, which are: (a) prediction of drug resistance levels in non-small-cell lung carcinoma treatments based on proteinligand interactions, (b) exploration of protein-protein cooperativity in POU/HMG/DNA complexes within human embryonic stem cells, (c) identification of protein-protein interfacial residues depending on protein sequential and spatial properties, and (d) detection of cooperativity between the periodic dinucleotides in a nucleosomal DNA molecule. Among these studies, molecular dynamics simulations and normal mode analysis are the two fundamental tools used to reveal the dynamic properties of molecules. The one- and three-dimensional structural properties of molecules were uncovered through a series of modeling and learning techniques. These were used to characterize the target molecular interactions or the interaction sites for analytical and statistical studies. Specifically, EGFR mutants-inhibitor interactions were studied in (a), and the binding free energy of each system was exacted based on molecular dynamics simulations. For a group of clinical subjects, a combination of their EGFR mutation features and personal features led to a promising personalized model for predicting the levels of drug resistance. In (b), normal mode analysis was implemented to reveal the motions of Oct and Sox proteins, and these motions were comparably investigated to reveal valuable Oct-Sox cooperative modes in embryonic stem cells. Protein-protein interfacial residues were characterized using protein sequential and spatial properties in (d). Protein interface prediction models were constructed relying on a number of learning mechanisms, while marginal performance differences between them implied an insensitivity of these structural data to different learning machines. In (d), intra-molecular interactions within a nucleosomal DNA molecule in normal mode motions were examined. Periodic dinucleotides were found generally located at the peaks or valleys of nucleosomal DNA motions, and a frequently-occurred dinucleotide pattern was uncovered. Overall, these works will contribute to many practical fields such as cancer drug discovery, disease research, and the design of specialized therapies.

Date of Award	3 Oct 2014
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Hong YAN (Supervisor)