Performance of 3D-Database Molecular Docking Studies into Homology Models

The performance of docking studies into protein active sites constructed by homology model building was investigated using CDK2 and factor VIIa screening data sets. When the sequence identity between model and template near the binding site area is greater than approximately 50%, roughly 5 times more active compounds are identified than would be found randomly. This performance is comparable to docking to crystal structures.