[ Search | Browse Families | Software | Information | Home ]
The central element of HOMSTRAD is a collection of structure-based alignments for individual families. Although we use automatic procedures such as MNYFIT (Sutcliffe et al., 1987), STAMP (Russell & Barton, 1992) and COMPARER (Sali & Blundell, 1990; Zhu et al., 1992) to align 3D structures, the results have been examined individually and manual adjustments have been made if necessary.
One of the problems in dealing with protein structures is that they are so much more complex than sequences and it is sometimes difficult to extract useful information. In HOMSTRAD, we use the program JOY (Overington et al., 1990; Mizuguchi et al., 1998) to overcome this problem. JOY was developed to highlight important structural features in a structure-based sequence alignment so that one can easily identify structurally important positions. Conserved structural features revealed by JOY can often define "structural fingerprints" for the family. These features are useful in producing accurate sequence-structure alignments and even identifying distant homologues whose structures are not known.
This is more systematically exploited by the homology recognition program FUGUE. It uses environment-specific substitution tables that were derived from the HOMSTRAD alignments and structure-dependent gap penalties, in order to associate query sequences with homologues with known structure. FUGUE has been shown to be more sensitive than PSI-BLAST and among the best methods for fold recognition. It is, thus, worth performing a FUGUE search against HOMSTRAD even if your protein appears to show no sequence similarity to any structures.
If you already know that your protein belongs to one of the known families, you can still use FUGUE to produce a reliable alignment between your sequence and the structures in the family (by clicking the blue 'align' icon). The generated alignment can be directly read in to modelling software such as MODELLER.
Created: 1 Aug 1997
Last modified: 23 Aug 2002