Deriving a receptor based QSAR model using docking data

It has long been a problem with all docking routines to find a scoring function that optimizes not only the chances of successfully predicting the binding mode of a putative binder, but which also tries to broadly reflect the potential binding affinity of the ligand with the modelled receptor. When the many approximations made when performing any docking run are taken into account – often of a very coarse nature in order to enable the calculation to be made at all -, it is no surprise that trying to make this correlation is not usually significantly helpful. However, the potential reward if binding affinities could be derived from docking scores is so great that there is never any shortage of attempts or claims to do so. Alternatively, if a significant amount of related SAR data is already known about a system about whose binding mode we might be reasonably confident, then 3Dstructural information derived from docking is quite easy to incorporate into a Quantitative Structure activity relationship (QSAR) and this can be significantly and semi-quantitatively helpful when considering future modifications to related structures.