This talk will address several related questions. When structural information is available how can we use it to design active compounds? What methods should be used? What feed-back is helpful? How can this be reduced to practice in a drug discovery project operating under strict time constraints?

A suit of programs, developed by the authors, for designing and evaluating drugs interactively at the computer terminal will be presented. These programs do not predict potency but enable compounds to be designed which have a good ``fit'' to the binding site or a good ``match'' to a pharmacophore or existing set of active molecules. Our experience has been that, being able to rapidly detect poor compounds and see what is wrong with them, can lead to a design cycle where up to 50 compounds may be modeled for each one synthesized.

The program reports within a few seconds, an estimated molecular mechanics association energy of a fully minimized structure, along with the energy of the ligand above its global minimum energy int he free state. The structure can be visualized superimposed on a pharmacophore model or in a binding site. A binding site surface can be displayed which makes it easy to see atoms which are poorly placed. In addition, individual internal ligand torsional angles and non-bonded interactions with high energy can be high-lighted.

One of the advantages of this approach is that it is possible to depart further from the core structure of a lead compound, thus accessing a wider variety of ligands. This may lead to targets that are easier to synthesize and/or have superior bioavailability prospects.

It is clear that being able to estimate potency with reasonable certainty could be very helpful since it would narrow down the range of compounds to be synthesized and give greater confidence to the chemist embarking on a novel and perhaps difficult synthesis.

Progress in potency prediction will be briefly reviewed, along with some theoretical background. Questions for discussion will be proposed.

This is joint work with Regine Bohacek.

Back to Workshop Schedule

1996-1997 Mathematics in High Performance Computing