Combination of Structural based Pharmacophore and Docking Virtual Screens: An Efficient Structure-based Protocol for Lead Identification
Filip Fratev
Micar21 Ltd., Persenk Str. 34B, 1407 Sofia, Bulgaria
Conference: ACS SWRM, October 29 – November 1At: Lubbock, TX, USA
Abstract
The structural and ligand base In silico screens are of great help during the drug discovery nowadays and their combination has been shown to provide better results than using a single approach. However, for many targets the number of known compounds is insufficient for traditional ligand base pharmacophore screen thus making this combination impossible. In a current study we tested whether the combination of recently introduced structural based pharmacophore approach with docking methodology, followed by rescoring with the new GlideSP scoring function, can achieve reasonable virtual screening (VS) results. These methods were tested also alone. All calculations were based on the full DUD-E benchmark set (102 targets). Surprisingly, we found that the structure based pharmacophore hypothesis created by the docking more than 650 fragments performs well for most of the targets with an average active ligand recovering rate of 10-15% during the first 10% database screen. The MD prepared structures future increased these values. These results are equal to the ligand based pharmacophore VS. The docking technique performed much better. Early enrichment performance shows on average that about 30% of all known actives are recovered in screening the top-ranked 1% of recovered decoys, similarly to the previous DUD tests. However, the main aim of any real drug discovery project is the very early enrichment performance, i.e. how many active compounds we will be discovered between the first 10 to 50 ligands, which we named here efficiently. According to or data the combination of these methods achieve an impressive result of nearly 75% success rate/efficiently for all 102 targets, even reaching often 100% for the top 10 ligands. Based on our data we developed a new VS protocol. We use the top 10% of ligands found by the structural based pharmacophore search as an input to the docking and scoring by GlideSP. This workflow was also used in a real live drug discovery projects searching for new Gly2 and AKT1 inhibitors. As a result we achieved over 70% success discovery rates. Moreover, discovered novel classes of inhibitors had both good selectivity and activity in a nanomolar range. Thus, we hope that our approach will be in help during the drug discovery projects and in particular in a case when no any know ligands are available for the studied protein.