|
Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT
| ABSTRACT - Internet Electron. J. Mol. Des. September 2005, Volume 4, Number 9, 659-670 |
Structure Prediction of Segments with Low Target-Template Similarity
in Comparative Protein Modeling Using a Reduced Protein Model
Andrzej Szymoszek and Martin Zacharias
Internet Electron. J. Mol. Des. 2005, 4, 659-670
|
Abstract:
Theoretical prediction of protein structures is important because
the number of sequenced proteins grows much faster than the
number of experimentally determined 3D structures. Among
theoretical methods, homology or comparative modeling of
unknown 3D protein structures (targets) has been established. It
is based on experimental structures of proteins (templates) with
sequence similarity to the target. The method is, however,
limited by the degree of sequence identity. Frequently, the target-template
sequence alignment is non-uniform along the sequence.
In the present study the possibility to model segments of low
target-template similarity by a systematic conformational search
based on a reduced protein model has been explored. The force
field is based on the concept of residue-residue contact energies
and allows to generate a large number of putative conformations
by energy minimization and selection of favorable
conformations. The approach was tested on a protein of known
structure by splitting the protein into mobile and
conformationally restrained regions. The mobile regions
represented putative regions with no structural information from
a template (the conformationally restrained regions represented
segments that can be modeled accurately based on a template).
The residue-based reduced protein model does not allow accurate
structure prediction of a complete protein. However, our results
demonstrate that with the test system and the present method it is
possible to successfully pick out segment topologies close to
experiment among a variety of possible structures, if the rest of
the protein structure is accurately defined. The approach could be
useful in comparative modeling in cases where most of the target
protein can be modeled accurately except for segments (beyond
the length of a loop) for which no template structure is available.
Randomly generated protein segment structures are subjected to
energy minimization employing a reduced protein model and
using positional restraints for conserved parts of the protein
structure as well as distance constraints to enforce a preset
secondary structure. The alpha-helical test protein results are
compared to the experimental protein structure. There is a
correlation between energy of a reduced protein structure, and its
similarity to the experimentally known structure, evaluated by
the root mean square deviation (rmsd) of corresponding atoms.
Low energy structures can be pre-selected for further refinement.
Our reduced protein modeling approach has been developed as a
possible tool to improve homology modeling in regions of low
target-template sequence similarity. Although the initial tests of
the model on a mainly alpha-helical structure showed quite
reasonable performance, further testing of the model is required
to make this approach generally applicable.
|