Strategies of Non-Sequential Protein Structure Alignments


Aysam Guerler (guerler@chemie.fu-berlin.de)
Ernst-Walter Knapp (knapp@chemie.fu-berlin.de)

Freie Universität Berlin, Institute of Chemistry and Biochemistry, Fabeckstrasse 36a, 14195 Berlin, Germany


Abstract

Due to the large number of available protein structure alignment algorithms, a lot of effort has been made to define robust measures to evaluate their performances and the quality of generated alignments. Most quality measures involve the number of aligned residues and the RMSD. In this work, we analyze how these two properties are influenced by different residue assignment strategies as employed in common non-sequential structure alignment algorithms. Therefore, we implemented different residue assignment strategies into our non-sequential structure alignment algorithm GANGSTA+. We compared the resulting numbers of aligned residues and RMSDs for each residue assignment strategy and different alignment algorithms on a benchmark set of circular-permuted protein pairs. Unfortunately, differences in the residue assignment strategies are often ignored when comparing the performances of different algorithms. However, our results clearly show that this may strongly bias the observations. Bringing residue assignment strategies in line can explain observed performance differences between entirely different alignment algorithms. Our results suggest that performance comparison of non-sequential protein structure alignment algorithms should be based on the same residue assignment strategy.

[ Full-text PDF |Table of Contents ]


Japanese Society for Bioinformatics