Phyre2 - Protein Homology/analogY Recognition Engine V 2.0 Webserver for protein structure prediction based on SCOP and PDB, and HMM-HMM alignment techniques for remote homology detection.
Phyre2 predicts the 3D structure of a protein sequence based on HMM-HMM alignment techniques. For a given sequence, it detects known homologues based on PSI-Blast, constructs a hidden Markov model (HMM) of the sequence based on the detected homologues and scans this HMM against a database of HMMs of known protein structures. The database is based on SCOP and latest depositions in PDB. The top 20 highest scoring matches of the query to known template structures are used to construct 3D models of the query. Here, the subset of the top 20 matches with confidence > 50% is being converted to a unified annotation. The list is traversed in decreasing order of confidence and non-overlapping templates are added in the annotation. Matches that have significant overlap (> 30 residues or > 50% of the length of the smallest template) with previously selected matches are discarded. Therefore, Phyre2 results as presented here are not necessarily in agreement with the original results.
Group Leader:Michael J E Sternberg, Lawrence A Kelley