Gergely Csaba1, Fabian Birzele, Ralf Zimmer. 1. Department of Informatics, Practical Informatics and Bioinformatics Group, Ludwig-Maximilians-University, Amalienstr 17, D-80333 Munich, Germany.
Abstract
MOTIVATION: Protein structure comparison exhibits differences and similarities of proteins and protein families and may help to elucidate protein sequence and structure evolution. Despite many methods to score protein structure similarity with and without flexibility and to align proteins accurately based on their structures, a meaningful evolutionary distance measure and alignment method which models the cost of mutations, insertions and deletions occurring in protein sequences on the structure level is still missing. RESULTS: Here, we introduce a new measure for protein structure similarity and propose a novel method called phenotypic plasticity method (PPM) which explicitly tries to model the evolutionary distance of two proteins on the structure level by measuring the cost of 'morphing' one structure into the other one. PPM aligns protein structures taking variations naturally observed in groups of structures ('phenotypic plasticity') into account while preserving the overall topological arrangement of the structures. The performance of PPM in detecting similarities between protein structures is evaluated against well-known structure classification methods on two benchmark sets. The larger set consists of more than 3.6 million structure pairs from the SCOP database which are also consistently classified in CATH. In the current parameterization, PPM already performs comparable or better than other methods such as TM-Align and Vorolign on those two sets according to various evaluation criteria showing that the method is able to reliably classify known protein structures, to detect their similarities and to compute accurate alignments despite phenotypic plasticity. AVAILABILITY: Executables are available upon request. Datasets and supplementary data (datasets and superpositions) can be accessed on http://www.bio. ifi.lmu.de/PPM.
MOTIVATION: Protein structure comparison exhibits differences and similarities of proteins and protein families and may help to elucidate protein sequence and structure evolution. Despite many methods to score protein structure similarity with and without flexibility and to align proteins accurately based on their structures, a meaningful evolutionary distance measure and alignment method which models the cost of mutations, insertions and deletions occurring in protein sequences on the structure level is still missing. RESULTS: Here, we introduce a new measure for protein structure similarity and propose a novel method called phenotypic plasticity method (PPM) which explicitly tries to model the evolutionary distance of two proteins on the structure level by measuring the cost of 'morphing' one structure into the other one. PPM aligns protein structures taking variations naturally observed in groups of structures ('phenotypic plasticity') into account while preserving the overall topological arrangement of the structures. The performance of PPM in detecting similarities between protein structures is evaluated against well-known structure classification methods on two benchmark sets. The larger set consists of more than 3.6 million structure pairs from the SCOP database which are also consistently classified in CATH. In the current parameterization, PPM already performs comparable or better than other methods such as TM-Align and Vorolign on those two sets according to various evaluation criteria showing that the method is able to reliably classify known protein structures, to detect their similarities and to compute accurate alignments despite phenotypic plasticity. AVAILABILITY: Executables are available upon request. Datasets and supplementary data (datasets and superpositions) can be accessed on http://www.bio. ifi.lmu.de/PPM.