Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

Open Access Software

Integrated web service for improving alignment quality based on segments comparison

Dariusz Plewczynski12*, Leszek Rychlewski1, Yuzhen Ye3, Lukasz Jaroszewski4 and Adam Godzik34

Author Affiliations

1 Bioinformatics Laboratory, BioInfoBank Institute, Poznan, Poland

2 Interdisciplinary Centre for Mathematical and Computational Modeling, University of Warsaw, Poland

3 The Burnham Institute, La Jolla, USA

4 Bioinformatics Core JCSG, University of California San Diego, La Jolla, USA

For all author emails, please log on.

BMC Bioinformatics 2004, 5:98  doi:10.1186/1471-2105-5-98

Published: 22 July 2004

Abstract

Background

Defining blocks forming the global protein structure on the basis of local structural regularity is a very fruitful idea, extensively used in description, and prediction of structure from only sequence information. Over many years the secondary structure elements were used as available building blocks with great success. Specially prepared sets of possible structural motifs can be used to describe similarity between very distant, non-homologous proteins. The reason for utilizing the structural information in the description of proteins is straightforward. Structural comparison is able to detect approximately twice as many distant relationships as sequence comparison at the same error rate.

Results

Here we provide a new fragment library for Local Structure Segment (LSS) prediction called FRAGlib which is integrated with a previously described segment alignment algorithm SEA. A joined FRAGlib/SEA server provides easy access to both algorithms, allowing a one stop alignment service using a novel approach to protein sequence alignment based on a network matching approach. The FRAGlib used as secondary structure prediction achieves only 73% accuracy in Q3 measure, but when combined with the SEA alignment, it achieves a significant improvement in pairwise sequence alignment quality, as compared to previous SEA implementation and other public alignment algorithms. The FRAGlib algorithm takes ~2 min. to search over FRAGlib database for a typical query protein with 500 residues. The SEA service align two typical proteins within circa ~5 min. All supplementary materials (detailed results of all the benchmarks, the list of test proteins and the whole fragments library) are available for download on-line at http://ffas.ljcrf.edu/darman/results/ webcite.

Conclusions

The joined FRAGlib/SEA server will be a valuable tool both for molecular biologists working on protein sequence analysis and for bioinformaticians developing computational methods of structure prediction and alignment of proteins.

Keywords:
Library of protein motifs; Profile-profile sequence similarity (BLAST; FFAS); Fragments library (FRAGlib); Predicted Local Structure Segments (PLSSs); Segment Alignment (SEA); Network matching problem