Figure 1
From: Regression applied to protein binding site prediction and comparison with classification
Method pipeline overview. Overview of the method pipeline. The input data are protein 3D conformations in PDB format. The surface of the proteins are computed and some properties are extracted. In the case of training proteins, the binding sites are identified. The surface are divided into patches which are associated with properties, and, in the case of training proteins, with the indices of the overlap with the binding site. The statistical model is trained and is used to give overlap scores to patches of application proteins. These scores are used to predict the locations of a potential binding sites.