Thesis (Ph.D.)--The University of Texas at Austin, 1999
Computer-assisted drug design (CADD) methods have made significant contributions to the overall drug development process. In this dissertation, we present three distinct projects, all of which contribute to the continued growth of CADD methodologies
Various CADD-related tasks require or are facilitated by the identification of those atoms “on the surface” of biological macromolecules. One obvious example is the computer graphic visualization and real-time manipulation of macromolecular surfaces. Other examples include algorithms which examine macromolecular surfaces looking for potential binding sites and software for modeling receptor-ligand interactions in terms of both the intermolecular interaction energies and the free energy of desolvation of the receptor site. In the first project, we describe a novel approach for the identification of those atoms on the surface of macromolecules. Our approach is based on the atomic contributions to the solvent-accessible surface area of macromolecules. As such, it is referred to as the Solvent-Accessible Surface Approach
The second project is related to novel, low-dimensional chemistry-space metrics referred to as the BCUT metrics. Briefly, the BCUT metrics are defined by the lowest and highest eigenvalues of a specially constructed matrix which incorporates both connectivity information and atomic properties relevant to drug-receptor interactions. The atomic properties are placed on the diagonal matrix elements and the connectivity information is placed on the off-diagonal matrix elements. A scaling factor is needed to balance these two types of information. We developed an approach to determine ranges of acceptable scaling factors for all BCUT metric types and we validated the BCUT values as true descriptors of molecular structure by using the Activity-Seeded, Structure-Based Clustering algorithm developed by Pearlman and Smith
Finally, a novel, semi-empirical approach for the General treatment of Solute-Solvent Interactions (GSSI) was developed to enable the prediction of solution-phase properties (e.g., free energy of desolvation, partition coefficient, and membrane permeability). Our GSSI approach is based on the notion that all solution-phase processes can be modeled in terms of one or more gas-to-solution transfer processes. Each gas-to-solution transfer process is calculated as a sum of the free energy of cavity formation (<math> <f> <g>D</g>G<sup>&j0;</sup><inf>cav</inf></f> </math>) and the free energy of solute-solvent interactions (<math> <f> <g>D</g>G<sup>&j0;</sup><inf>ssi</inf></f> </math>). <math> <f> <g>D</g>G<sup>&j0;</sup><inf>cav</inf></f> </math> is modeled in terms of the total solvent-accessible surface area of the solute. In modeling <math> <f> <g>D</g>G<sup>&j0;</sup><inf>ssi</inf></f> </math>, we take into account not only the energetic contributions from electrostatic, polarization and dispersion interactions but also entropic contributions from both the solute and the solvent. We demonstrate the utility of our GSSI approach by applying it to the prediction of free energies of gas-to-aqueous solution transfer, free energies of gas-to-hexadecane transfer, and octanol-water partition coefficients
