Shape-based virtual screening

In the validation study by Hawkins et al. (60), the shapematching method ROCS was compared to 7 well-known docking tools, in terms of their abilities to recover known ligands for 21 different protein targets. The comparative study showed that the 3D shape method (ROCS) performed at least the same as, and often better than, the docking tools studied. Their work indicated that shape-based virtual screening method could be both efficient (in terms of the computational speed) and effective (in terms of hit enrichment) in virtual screening projects. 

Kirchmair, J., Distinto, S., Markt, P., Schuster, D., Spitzer, G. M., Liedl, K. R, Wolber, G. (2009) How to optimize shape-based virtual screening choosing the right query and including chemical information. J Chem Inf Model49, 678-692. 

The shape-based virtual screening protocols performed better than the docking protocols overall. Finally, a consensus rank-byvote selection technique was used and the five best compounds were selected from the hits for synthesis and for determining antiviral activity. Remarkably, two of the five compounds (Fig. 13) showed activity at low micromolar ranges (0.022 pg/mL and 0.058 pg/mL, respectively). 

Proschak E et al (2007) Shapelets possibilities and limitations of shape-based virtual screening. J Comput Chem 29(1) 108-114... 

Structural alignments (ligand-based virtual screening) on the basis of a known active compound, molecules that show a good superimposition in shape and physicochemical features are searched for. 

Ballester, P. J., Finn, P. W., Richards, W. G. (2009) Ultrafast shape recognition evaluating a new ligand-based virtual screening technology. J Mol Graph Model 27, 836-845. 

The structure based virtual screening process includes also the use of some keywords. With these keywords FLAP fdters out matches and keeps them only if they make sense in terms of binding site shape. FLAP can also allow additional binding site volume (cavity expansion, useful when the protein structure under investigation is an homology model) and with the use of regions (definition of a sphere within each pharmacophore needs to have at least one point) or selection of a probe (enforcing a particular feature to be present in the calculated pharmacophores) certain constraints can also be added. 

If the target of the ligands under investigation is known, another possible approach is to compare ligands using the shape of the protein as a shape constraint and features in the protein cavity as additional constraints. As in the case of structure based virtual screening, keywords are used such as regions to define a sphere within which each pharmacophore needs to have at least one point, and the selection of a particular probe. 

A further important option for ligand-based virtual screening is to perform shape comparisons that have a pronounced scaffold hopping potential. An often used method of this kind is ROCS [31] (OpenEye Scientific Sofiware, 9 Bisbee Court Suite D, Santa Fe, NM 87508, USA. Available at http //www.eyesopen.com, March 20, 2009). ROCS employs continuous functions that are derived from atom-centered Gaussians to calculate the volume overlap between two 3D structures. The use of Gaussians drastically speeds up the computational process, and ROCS is able to search even databases with millions of compounds for molecules that can adopt shapes similar to the reference compound. As in the case of 3D pharmacophore searches, the low-energy conformations must be precomputed for the search database. 

---