Premium
Docking without docking: ISEARCH—prediction of interactions using known interfaces
Author(s) -
Günther Stefan,
May Patrick,
Hoppe Andreas,
Frömmel Cornelius,
Preissner Robert
Publication year - 2007
Publication title -
proteins: structure, function, and bioinformatics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.699
H-Index - 191
eISSN - 1097-0134
pISSN - 0887-3585
DOI - 10.1002/prot.21746
Subject(s) - docking (animal) , macromolecular docking , computer science , interface (matter) , biological system , similarity (geometry) , protein structure prediction , structural similarity , benchmark (surveying) , protein structure , protein–protein interaction , computational biology , algorithm , data mining , artificial intelligence , chemistry , biology , medicine , biochemistry , nursing , geodesy , bubble , maximum bubble pressure method , parallel computing , image (mathematics) , geography
The increasing number of solved protein structures provides a solid number of interfaces, if protein–protein interactions, domain–domain contacts, and contacts between biological units are taken into account. An interface library gives us the opportunity to identify surface regions on a target molecule that are similar by local structure and residue composition. If both unbound components of a possible protein complex exhibit structural similarities to a known interface, the unbound structures can be superposed onto the known interfaces. The approach is accompanied by two mathematical problems. Protein surfaces have to be quickly screened by thousands of patches, and similarity has to be evaluated by a suitable scoring scheme. The used algorithm (NeedleHaystack) identifies similar patches within minutes. Structurally related sites are recognized even if only parts of the template patches are structurally related to the interface region. A successful prediction of the protein complex depends on a suitable template of the library. However, the performed tests indicate that interaction sites are identified even if the similarity is very low. The approach complements existing ab initio methods and provides valuable results on standard benchmark sets. Proteins 2007. © 2007 Wiley‐Liss, Inc.