Computer Modeling of Biomolecular Systems

Stewart, James

Abstract

Software for full quantum mechanics modeling of biological macromolecules that is both fast and accurate is proposed. Cancer researchers could use this software to model protein and enzyme structure as well as enzyme reactions involving bond formation and breaking. The software would be based on a semi-empirical quantum chemical compute engine with a newly developed parameterization for all biological elements including the six principal transition elements. In this Phase I proposal, the question of whether an accuracy of 3 kcal/mol can be achieved for biological systems will be answered. If this answer is affirmative, the following tasks will be pursued in Phase II: A. Methods specific to proteins and other biomolecules would be developed. Existing methods are (a) too general, and (b) unsuitable for biochemical work, due to the lack of transition metals. By developing semiempirical methods that are specific to systems found in biochemistry, an increase in accuracy can be obtained, thus the average error in semiempirical methods should be decreased by about 60% relative to PM3. Similarly, limiting the range of properties optimized for transition metals to those found in biochemical systems, an accuracy at least equivalent to ab initio 6-31g would be achieved. B. Writing Graphical user interfaces (GUI) and utilities to allow reactions, transition states, UVVisible spectra, etc. to be modeled. These will allow users to run simulations using by issuing """"""""requests"""""""", which the GUI will convert into commands for the compute engine. The compute engine would run at a speed sufficient to allow real-time simulations to be performed. This tool will be useful for visualizing biochemical systems, and for understanding the mechanisms involved.