This application proposes work related to NIMH research topic number 1J. Commercialization by Tripos in both commercial diversity software and in collaborative drug discovery research is very likely. Combinatorial chemistry and high throughput screening provide results at remarkably high rates. However, individual data points are of low quality. Scientists must include their own judgment to prioritize synthesis suggestions from """"""""black boxes,"""""""" which compute diversity in abstract mathematical spaces of molecular metrics. To do so requires new methods to illustrate important structures in high dimensions by translating them to interpretable graphic displays. Drug discovery is based on the intuitive idea that similar compounds tend to have similar biological effects. A corollary of this """"""""neighborhood behavior"""""""" principle is that, if a metric space is valid, active molecules should be clustered within it. However, all metrics documented to be generally valid have roughly 1000 dimensions. The objective of the proposed research is to provide interactive 3D computer graphics displays which represent higher dimensional spaces by preserving the important local distances (nearest neighbor compound clusters). While the development will include heuristics specific to chemistry of drug-like organic compounds, the result may be useful for visualizing key aspects of arbitrary multidimensional data spaces.
Pharmaceutical companies currently buy Tripos software containing mathematical algorithms for diversity. They license the Optiverse screening library of compounds designed for optimal diversity. Tripos is also involved in collaborative drug discovery projects. Each of these current commercial applications will be improved by use of the visualization tools this project will produce. Potentially the results can be extended to create a new cheminformatics software package and serve as a basis for new bioinformatics offerings as well.
| Clark, R D; Patterson, D E; Soltanshahi, F et al. (2000) Visualizing substructural fingerprints. J Mol Graph Model 18:404-11, 527-32 |
