FBRHGRM Method 2_RankbyScore
3D QSAutogrid/R
A, C, e, HD, N, NA, OA probe s
3PCs
A ligand-based 3D QSAR prediction has been performed using the 3D QSAutogrid/R method.
The 3D QSAR model was previously composed using the structure-activity data available from the literature (considering only pure compounds with measured SPA binding affinity).
The binding poses of non co-crystallized ligands (used as training and test sets) have been refined by means of ensemble cross-docking (the 36 D3Rs' released FXR X-ray protein conformations have been included in the original ensemble of protein structures to enhance the docking resolution).
Cross-validation analysis of the refined 3D QSAR models suggested the probes A, C, e, HD, N, NA, OA as the most predictive, thus used to predict the activity of the D3R FXR agonists.
A rank by score protocol has been performed.
For each D3R FXR ligand an average pIC50 value is reported;
D3R's FXR ligands with multiple poses and selected from the docking calculations have been tested through the 3D QSAR. The final pIC50 of each ligand with selected multiple poses was then derived by weighting the specific pIC50-pose according to the relative pose occupancy.
In the SuppInfo folder is also included a csv file (Multiple_occupancies_pIC50s.csv) reporting the predicted pIC50s of the multiple considered poses, and the calculation applied to derive the relative corrected pIC50 value.