FINDSITEcomb: Drug Discovery Tool
In recent work, the performance of FINDSITEcomb was compared to several commercially and freely available docking programs against the DUD set – A Directory of Useful Decoys. We demonstrated that FINDSITEcomb has virtual screening accuracy better than the best docking method under the challenging condition that no templates > 30% sequence ID to the target are present in the ligand binding databases. We find an average area under the ROC curve, AUC, of 0.77 using crystal target structures and 0.74 using modeled structures with average TM-score ~ 0.75. The individual results for each DUD target can be viewed here. If we set the sequence identity cutoff to 95%, FINDSITEcomb will give the unprecedented mean AUC=0.90 using modeled structures. Thus, FINDSITEcomb does not require experimental structure to give the best performance. For a ~ 300 AA protein, FINDSITEcomb can screen ten million compounds within 2 days on a single computer node.
Notice: This server is freely available to all academic and non-commercial users.
Commercial users – to use this server, or request a download of all pre-computed scores, please send an email to Dr. Jeffrey Skolnick: skolnick@gatech.edu.
If you find this service useful, please cite the following paper: H. Zhou and J. Skolnick. 2013. FINDSITEcomb: A threading/structure-based, proteomic-scale virtual ligand screening approach. Journal of Chemical Information and Modeling. 53(1):230-240. PDF