PatchBag - a novel approach for quick detection of interface similarity and complementarity
Along with the massive expansion of the Protein Data Bank, the need for automatic ways to predict protein function is growing. Protein surfaces, and particularly intermolecular interfaces, encapsulate much of the information of the protein function. Thus, efficient methods for surface and interface comparisons are essential.
We present PatchBag – a novel approach to characterize protein surfaces and interfaces by the distribution of their overlapping surface patches. A surface patch is defined by a central surface carbon-α atom and its 4 nearest carbon-α atom (excluding the ones from the immediate neighboring residues along the protein backbone). In addition, we describe the patch orientation using a normal vector pointing away from the protein core. Interfaces are represented by considering only patches which their central residue is an interface residue (as defined by the 3D structure of a protein-protein complex).
PatchBag algorithm is highly efficient and can be applied on large datasets. We applied PatchBag to detect surface similarity in a dataset of 1,910 protein domains, and to identify pairing interfaces in a dataset of 1,514 interacting domains.
An illustration of extracting a PatchBag vector.