Voronoi site modeling: A computer model to predict the binding affinity of small flexible molecules.

Abstract

Voronoi site modeling is a novel method that simulates the binding of small flexible molecules to biological macromolecules. The model uses a linearized representation for a ligand molecule which accurately but compactly represents the allowed conformational space, allows accurate three dimensional structures to be generated efficiently, and realistically allows a ligand to choose the geometry and position giving the energetically most favorable interactions with the site model. The representation for the site model uses Voronoi polyhedra to specify the geometry of the site, and physicochemical properties for the energetics. The investigator supplies only the ligand structures and observed binding energies, along with a proposed site geometry. No additional assumptions are made about how the ligands bind, or which parts of the ligands are most important for binding. A collection of computer programs build a model to fit the observed binding energies without leaving outliers. This model predicts how each of the ligands binds in the site, and can be used to predict the strength and mode of binding for new compounds, regardless of chemical similarity to the original set of ligands. The computer programs which implement this model include many different algorithms from different disciplines. My contributions to this project have been primarily in the research, development, improvement and implementation of these algorithms. Nonlinear optimization techniques are applied to the problems of finding ligand conformations which fit into the site in an optimal way. Optimal site energetics are found using a nonsmooth optimization technique. A linear complementarity algorithm is used to locate important distances. Algorithms from computer science and database management are implemented to handle large amounts of data in an efficient manner. Voronoi polyhedra are located and manipulated using algorithms from computational geometry. The principles of software engineering are applied to help make the programs efficient, usable and maintainable.