Structural models in the assessment of protein druggability based on HTS data

Abstract

Insights on the potential of target proteins to bind small molecules with high affinity can be derived from the knowledge of their three-dimensional structural details especially of their binding pockets. The present study uses high-throughput screening (HTS) results on various targets, to obtain mathematical predictive models in which a minimal set of structural parameters significantly contributing to the hit rates or the affinity of the protein binding pockets for small molecular entities, is identified. An emphasis is given to focus on target variation aspect of the data by consideration of commonly tested compounds against the HTS targets. We identify ‘four-parameter’ models with R ², $R_{\text{adj}}^{2}$, SEE, and LOO q ² values of 0.70, 0.60, 0.27 and 0.50, respectively, or better. We demonstrate through cross-validation exercises that our regression models apply well on varied data sets. Thus we can use these models to estimate hit rates for HTS campaigns and thereby assign priority to drug targets before they undergo such resource intense experimental screening and follow-up.