Data Mining and Machine Learning Guided QM/MM and QM-Cluster Modeling of Enzymatic Reactions

点击翻译按钮获取中文摘要

Project Summary/Abstract Computational modeling methods have been widely applied in protein structure prediction, drug discovery and enzyme bioengineering to provide atomic-level insight into enzymatic reactions and functions. Accuracy and efficiency are the two goals that motivate the development of new methods in this field. However, the methodological best practices are still lacking in achieving high throughput and accuracy. In quantum mechanics/molecular mechanics (QM/MM) and QM-cluster enzyme modeling, series of decisions such as molecule partitioning into QM and MM regions, protonation states of residues, and computational setting rely on good understanding of the problem and knowledge of the enzyme as well as available computational methods. In this proposed project, machine learning methods will be applied in computational enzyme modeling for a better and more systematic solution. The proposed project is innovative as it combines a) data mining and machine learning on published experimental and computational works which will efficiently and systematically collect knowledge for research; b) machine learning methods can weigh different components of computational modeling and make optimal decisions automatically; c) the results of this work will provide a rational strategy for accurate and efficient QM/MM and QM-cluster simulations in future studies of different protein systems, drug design and even other scientific research domains. The proposed project will focus on two enzyme systems that will serve as case studies: a) Chorismate Mutase which is a potential target for designing antibiotics and b) the Cytochrome P450 superfamily of metalloenzymes which are largely involved in drug metabolism via various reaction mechanisms.