An Automatic Virtual Screening System

Notwithstanding well-known algorithmic weaknesses, molecular docking screens have had important successes in recent years, and are now the most practical technique to leverage structure for ligand discovery. Unfortunately, barriers to entry have largely restricted the techniques to experts and their collaborators. Docking databases are expensive to acquire, require considerable manipulation, and the software is byzantine. Many pitfalls await the unwary docker, such as the preparation of the protein binding site and the choice of dozens of parameters. This has diminished the impact of the technique and limited the sorts of problems to which it can be applied. To address this situation we are building tools and databases that will bring docking to a broad audience, in the spirit of BLAST, and allow its application to new questions. This talk will describe the development of and experience with an early prototype of an automated web-based docking system and its components. We will illustrate some of the strengths and weaknesses of this approach against targets of current or recent therapeutic interest.