A Structure Guided QSAR: A Rapid and Accurate technique to predict IC<sub>50</sub>: A Case Study

In this study, the role of highly structured water molecules present in the active site of the leukotriene A4 hydrolase (LTA4H) enzyme has been critically examined through the docking experiments. It was observed that different experiments were necessary to perform the docking studies. The ligands capable of displacing or interacting with bound-water(s) displayed different binding poses as well as the scores. The docking scores E (CvdW) from Glide and ChemScore (CS) from FlexX, with and without bound-waters, obtained through different docking experiments were used to construct two structure guided bi-parametric linear regression models using the IC50 enzyme activity data. The predictive squared correlation coefficients (Q2) obtained for these models, with and without bound waters, were found to be 0.73 and 0.67, respectively. These models were validated using test and validated sets of compounds. Utilizing these QSAR models, 409 proposed structures were docked and their respective predicted IC50 data were generated. From the in-silico evaluation of these 409 proposed structures representing diverse chemotypes, 39 compounds were triaged, synthesized and evaluated in the enzyme inhibition assay. The predicted and experimental biological data (IC50) was correlated and the square of the correlation coefficient (R2) between the observed and calculated IC50 was found to be 0.87. Keywords: Crystallographic waters, structure based QSAR, docking, scoring function.