Multiple binding sites

We provide experimental and computational information associated to a dataset of 262 protein chains of varied sizes and functions:

• A set of 329 functional interaction sites extracted from known complexes formed between these proteins


• A set of 370 interaction regions obtained by merging more than 20 000 functional interaction sites known for these proteins and their close homologs


• Residue propensities to be found at an interface, inferred from complete cross-docking of these proteins


• A set of interacting patches predicted at the surface of these proteins by using evolutionary information, geometry and docking-based propensities.

We also provide a computational tool, dynJET², for predicting interacting patches based either on protein sequence and structure analysis or on any pre-computed residue based property. dynJET² implements four scoring strategies aimed at detecting different types of protein interfaces:

• SCcons targets very conserved residues to form a seed which is then extended by conserved residues displaying physico-chemical properties often found at interfaces ; an outer layer of protruding and physico-chemically favorable residues is finally added.


• SCnotLig detects conserved residues that are not too buried (to avoid small-ligand binding pockets) ; these residues form the seed and the extension of the predicted patch ; an outer layer of protruding and physico-chemically favorable residues is finally added.


• SCgeom disregards evolutionary information and looks for protruding residues with favorable physico-chemical properties.


• SCdock exclusively uses a pre-computed property to define the seed, the extension and the outer layer. In the data reported here, the property consists in residue propensities to be found at interfaces, inferred from complete cross-docking calculations.

Instructions for installing and running the tool are available here.