Initially the service is available for compound profiles against 7 receptor models. Each model was trained using 2.3 million clean drug-like compounds downloaded from ZINC molecular database. The models were created using our modelling strategy given in paper "Efficient iterative Virtual Screening with Apache Spark and Conformal Prediction" accepted in Journal of Cheminformatics.
Once the profiles are ready, each compund can be docked against a particular receptor by selecting pdb code of available receptor and pressing the Run QVina button. Score will appear against each compound.
Quick Vina 2.0 has been used for docking. A ligand with low score generally means a higher affinity. Similarly, prediction of low-score or green means higher affinity and vise versa.
Tutorial: A tutorial for running CPVSAPI on a local system and creating custom docker images for new receptors is available on GitHub.
Predicting Target Profiles with Confidence as a Service using Docking Scores
Laeeq Ahmed, Hiba Alogheli, Staffan Arvidsson Mc Shane, Arvid Berg, Jonathan Alvarsson, Anders Larsson, Wesley Schaal, Erwin Laure and Ola Spjuth