Homology Modeling: Generating Structural Models to Understand Protein Function and Mechanism - Computational Modeling of Biological Systems - page 103

Fig. 3 Remote homolog detection. An outline of different strategies involved in detecting remote
homologs: sequence-based and structure-based methods and by using meta-servers. The steps
involved in each of these strategies are also outlined as a flowchart
in protein sequence [ 31 ]. HMM-based methods rely on constructing an HMM-
profile for any given sequence, based on a seed alignment generated either using
BLAST or manually. Most of these methods have thousands of such profiles for all
known sequences. Using these profiles, for any query sequence, sequence-profile
and profile-profile matching can be performed to identify significant structural
homologs. All known domains are usually arranged in hierarchical families based
on either function or fold to enable quick retrieval of matches. A given sequence
is searched against HMM-profiles of families that have at least one representative
structure, a process called sequence-profile alignment (Fig. 3 ). A logical expansion
of sequence-profile alignment is profile-profile alignment, where a profile is
constructed based on evolutionary conservation of the query sequence. The seed
alignment for constructing the profile for the query sequence is usually obtained
using PSI-BLAST. Once an HMM-profile is generated for the query sequence using
PSI-BLAST-based multiple sequence alignment, this profile is searched against
other profiles that have at least one representative structure. Apart from providing
a template structure for constructing a homology model, these profile-profile and
sequence-profile alignments provide a quick means to predict domain boundaries
and possible function of the sequence. For example, scanning the query sequence
using Pfam [ 21 ] (a database of HMM-profile based domain families) will identify
the different domains in the sequence as well as possible functional and structural
information of the identified domains.
Structure-based threading [ 32 - 35 ] forms the basis of the second group of
protocols (Fig. 4 ). We can observe high diversity in the specific protocol followed
by each structure-based threading program to identify remote homologs. Since each