27 August - 1st September, 2000. Duesseldorf, Germany, 13th European Symposium on Quantitative Structure-Activity Relationships

Fragment-based QSAR Method for High-throughput Calculations

Alanas Petrauskas, Remigijus Didziapetris

Abstract

New QSAR method is based on derivation of fragmental increments which depend on fragment's position in drug structure. It is similar to a Free-Wilson approach which has been widely used for small linear peptides. Now it is extended to any other chemical structures with variable branching, cyclization and spatial orientation of the radicals. The major difficulty is in reduction of high number of linear regression variables to a level of good statistical justification. This is achieved by establishing relations among binding energies in different drug sites as well as other empirical correlations. A model training set of 400 diverse HIV-1 protease inhibitors produced initially 319 independent variables which was then reduced to only 60. Final correlation produced N=375, R=0.986, S=0.26, F=195. The obtained accuracy is close to experimental. This method does not require any 3D structure optimization, so it can be used in high-throughput calculations.

Download the poster in PDF format (63,8 Kb) or ChemSketch format (110 Kb).