Correlation Between Virtual Screening (VS) Success and Druggability for Protein-Protein Interaction (PPI) TargetsLewis R. Vidler1, Nathan Brown1, Swen Hoelder1
|1Institute of Cancer Research|
|PPIs are involved in all biological pathways and are becoming increasingly common targets for drug discovery. One of the main challenges in targeting these interactions is that the binding sites have evolved to recognise protein binding partners, not small molecules. As a result of this, the binding sites are frequently large (>1000 Å2), featureless and solvent exposed,1 making the identification of small molecule inhibitors difficult. The varying difficulty in identification of small molecule inhibitors of a particular protein binding site has been termed its 'druggability'. Here we have used the same definition of a druggable protein as Hopkins et al.2: proteins able (or predicted to be able) to bind drug-like molecules.|
The work presented here covers the application of similar VS protocols to four PPI targets BRD4, HOP, SIAH1 and TLE1. Crystal structures of each of these proteins have been used for the application of structure-based VS methods. Pharmacophore similarity searches using the protein as an exclusion volume and Shape-based similarity searches using bound peptides or small molecules were used to screen two million commercially available molecules. The results of these were submitted to docking experiments and, following a set of filters designed to reduce the numbers of false positives, compounds were selected for purchase based upon these docked poses.
The druggability of each of these proteins has been assessed using SiteMap.3 To our knowledge this is the only method for which bound water molecules can be included as part of the assessed binding site. The four PPI targets demonstrated a wide range of druggability as assessed by SiteMap, from druggable (BRD4) to undruggable (SIAH1). These assessments are in line with other available data including HTS and fragment screens performed in-house, and literature data.
These druggability assessments correlated well with the observed success of the virtual screens. For BRD4, 143 compounds were selected using our VS approach, purchased and screened. From those, 10 novel inhibitors were identified corresponding to a hit rate of 7% and outperformed a control experiment that failed to identify genuine inhibitors. At the other end of the druggability scale, 101 compounds were selected, purchased and screened against SIAH1. None of these compounds demonstrated genuine inhibition of the PPI consistent with the druggability assessment and the lack of success from other screening approaches. Interestingly, HOP demonstrated intermediate druggability and this is reflected in the screening efforts. HTS and fragment screens failed to reveal any tractable hits, whereas the VS was able to identify genuine inhibitors (protein NMR confirmation assay), albeit with compounds that lie outside what would currently be considered drug-like.
This work demonstrates the ability of VS to identify genuine inhibitors of more tractable PPI targets and in one case probe chemical space not accessed by a more limited screening collection and identify hit matter. Also this is a prospective validation of SiteMap, exemplifying its discrimination between PPIs of varying druggability and its use as a predictor of virtual screening success.
 Conte, L. L.; Chothia, C.; Janin, J. The Atomic Structure of Protein-Protein Recognition Sites. J. Mol. Biol. 1999, 285, 2177-2198.
 Hopkins, A. L.; Groom, C. R. The druggable genome. Nat. Rev. Drug Discovery 2002, 1, 727–730.
 Halgren, T.,A., Identifying and Characterizing Binding Sites and Assessing Druggability, J. Chem. Inf. Model. 2009, 49, 377-389.