Empirical scoring functions. II. The testing of an empirical scoring function for the prediction of ligand-receptor binding affinities and the use ofBayesian regression to improve the quality of the model

This paper tests the performance of a simple empirical scoring function on a set of candidate designs produced by a de novo design package. The scorin g function calculates approximate ligand-receptor binding affinities given a putative binding geometry. To our knowledge this is the first substantial test of an empirical scoring function of this type on a set of molecular d esigns which were then subsequently synthesised and assayed. The performanc e illustrates that the methods used to construct the scoring function and t he reliance on plausible, yet potentially false, binding modes can lead to significant over-prediction of binding affinity in bad cases. This is antic ipated on theoretical grounds and provides caveats on the reliance which ca n be placed when using the scoring function as a screen in the choice of mo lecular designs. To improve the predictability of the scoring function and to understand experimental results, it is important to perform subsequent Q uantitative Structure-Activity Relationship (QSAR) studies. In this paper, Bayesian regression is performed to improve the predictability of the scori ng function in the light of the assay results. Bayesian regression provides a rigorous mathematical framework for the incorporation of prior informati on, in this case information from the original training set, into a regress ion on the assay results of the candidate molecular designs. The results in dicate that Bayesian regression is a useful and practical technique when re levant prior knowledge is available and that the constraints embodied in th e prior information can be used to improve the robustness and accuracy of r egression models. We believe this to be the first application of Bayesian r egression to QSAR analysis in chemistry.