TASSER-Lite: An automated tool for protein comparative modeling
Shashi Bhushan Pandit 1, Yang Zhang 2 and Jeffrey Skolnick 1*
1 Georgia Institute of Technology
2 The University of Kansas
* To whom correspondence should be addressed. E-mail: skolnick{at}gatech.edu.
Submitted on March 1, 2006
Revised on May 11, 2006
Accepted on 22 August 2006
 |
Abstract |
|---|
The present study involves the development of a rapid comparative modeling tool for homologous sequences by extension of TASSER methodology, developed for tertiary structure prediction. This comparative modeling procedure was validated on a representative benchmark set of proteins in the Protein Data Bank (PDB) composed of 901 single domain proteins (41-200 residues) having sequence identities between 35-90% with respect to the template. Using a Monte Carlo, MC, search scheme with the length of runs optimized for weakly/non-homologous proteins, TASSER often provides appreciable improvement in structure quality over the initial template. However, on an average, this requires ~29 hours of CPU time per sequence. Since homologous proteins are unlikely to require the extent of conformational search as weakly/nonhomologous proteins, TASSER's parameters were optimized to reduce the required CPU time to ~17 min, while retaining TASSER's ability to improve structure quality. Using this optimized TASSER (TASSER-Lite), we find an average improvement in the aligned region of ~10% in root-mean-square deviation, RMSD, from native over the initial template. Comparison of TASSER-Lite with widely used comparative modeling tool MODELLER showed that TASSER-Lite yields final models that are closer to the native. TASSER-Lite, is provided on the web at (http://cssb.biology.gatech.edu/skolnick/webservice/tasserlite/index.html).
Key Words:
protein modeling, protein tertiary structure, threading
