My current research efforts are focussed on the problem of designing drug molecules (inhibitors) against highly mutable targets such as HIV protease. As part of an NIH-funded collaboration with industrial and academic partners, I have been developing a computer-based drug design system called ALMS (Analysis of Ligand binding with Multiple Substitutions). ALMS works by attaching user-defined selections of chemical fragments to selected sites on a pre-existing "framework" compound, and rapidly orienting the attached fragments in the binding site of a target protein using a genetic algorithm. ALMS works "combinatorially," in that it builds all possible combinations of selected fragments at framework sites, and can easily generate many thousands of compounds.

The drug compounds thus generated can be "screened" by evaluating the interaction energy between each inhibitor and the protein binding site. This is done using a molecular mechanics potential energy function. Compounds with lower (more negative) interaction energy are predicted to be better binders, more effective inhibitors, and better candidates for synthesis and testing. Using ALMS it is also possible to correlate observed biological activity with energy, and to construct models that directly predict measured drug efficacy.

We are also working on the important question of including the effects of solvation in estimates of drug binding energy. This is being done using continuum models that I have developed, which take into account the charge distribution of the solute and the geometry of the molecular surface to predict the effects of polarized solvent. This technique is now being integrated with ALMS.