Aiming at an efficient and generalizable computational framework for 3D image retrieval, in this project, we are developing effective computational methods with deep learning algorithms at the heart, that can compare protein 3D structural data to identify specific drug (ligand) binding
sites. We use multi-channel 3D images of voxelized protein structure as input, where crucial information about the proteins, such as 3D shape and physicochemical properties, are included in the images.