BioDrugScreen: a computational drug design resource for ranking molecules docked to the human proteome

Abstract

BioDrugScreen is a resource for ranking molecules docked against a large number of targets in the human proteome. Nearly 1600 molecules from the freely available NCI diversity set were docked onto 1926 cavities identified on 1589 human targets resulting in >3 million receptor-ligand complexes requiring >200,000 cpu-hours on the TeraGrid. The targets in BioDrugScreen originated from Human Cancer Protein Interaction Network, which we have updated, as well as the Human Druggable Proteome, which we have created for the purpose of this effort. This makes the BioDrugScreen resource highly valuable in drug discovery. The receptor-ligand complexes within the database can be ranked using standard and well-established scoring functions like AutoDock, DockScore, ChemScore, X-Score, GoldScore, DFIRE and PMF. In addition, we have scored the complexes with more intensive GBSA and PBSA approaches requiring an additional 120,000 cpu-hours on the TeraGrid. We constructed a simple interface to enable users to view top-ranking molecules and access purchasing and other information for further experimental exploration.

Figure 1.

A snapshot of the BioDrugScreen page with options to rank ligands that were pre-docked to a large number of cavities of receptors from the human proteome.

Figure 2.

A snapshot depicting the three search options available in BioDrugScreen. (A) A pull-down menu lists all the targets within BioDrugScreen. The user then selects the ‘Chain’ and ‘Binding Cavity’ pull-down menus to select chain and cavity; (B) targets can be found through a keyword search. (C) Targets can be identified through a BLAST sequence search. The user pastes the sequence within the box provided to conduct the search.

Figure 2.

A snapshot depicting the three search options available in BioDrugScreen. (A) A pull-down menu lists all the targets within BioDrugScreen. The user then selects the ‘Chain’ and ‘Binding Cavity’ pull-down menus to select chain and cavity; (B) targets can be found through a keyword search. (C) Targets can be identified through a BLAST sequence search. The user pastes the sequence within the box provided to conduct the search.

Figure 3.

Snapshot of BioDrugScreen following the selection of a receptor through one of the three mechanisms provided. A Jmol session depicts the 3D structure of a molecule bound to the protein Caspase-9. The molecule is shown in capped-sticks representation with atoms color-coded based on atom types (N, C, S and H in blue, grey, yellow and white, respectively). The protein is shown in wire rendering. A table below the Jmol session provides a listing of the molecules docked to the target and ranked using a series of standard scoring functions along with various links for further structural analysis and information from other sites.