A novel method to three-dimensional-QSAR using genetically evolved molecular field data and artificial neural networks

Computer Aided Drug Design (CADD) can be divided into two categories: (1) receptor based design (where the structure of the active site has been elucidated) and, (2) ligand based design (where the structure of the active site has not been elucidated). Since the vast majority of drug targets have not had their structure elucidated it seems reasonable to study the better approach.; The most common tool in a ligand based design is a three dimensional quantitative structure activity relationships (3D-QSAR), of which the most widely used method is the Comparative Molecular Field Analysis (CoMFA) method developed by Richard Cramer III of Tripos Associates. A CoMFA study aims to correlate the biological activity of a series of compounds with certain molecular properties by means of a regression analysis like the partial least squares analysis (PLS). The largest drawback with this approach is the fact that a PLS analysis can only make linear correlations. However, it is conceivable, if not often probable, that the complex relationship between molecular structure and biological activity is not linear. The goal of this research is to address the inherent problems in forcing a linear relationship upon the development of a quantitative structure activity relationship (QSAR) by using parallel genetic algorithms coupled with non-linear neural networks.