A novel complexity measure for comparative analysis of protein sequences from complete genomes

Abstract

Analysis of sequence complexities of proteins is an important step in the characterization and classification of new genomes. A new measure has been proposed to compute sequence complexity in protein sequences based on linguistic complexity. The algorithm requires a single parameter, is computationally simple and provides a framework for comparative genomic analysis. Protein sequences were classified into groups of high or low complexity based on a quantitative measure termed F_c, which is proportional to the fraction of low complexity sequence present in the protein. The algorithm was tested on sequences of 196 non-homologous proteins whose crystal structures are available at ≤2.0 Å resolution. Protein sequences of high complexity had 'globular' structures (95% agreement), whereas those of low complexity had non-globular structures (80% agreement). Application of this measure to proteins of unknown structure/function from different genomes revealed that the sequences of high complexity constitute the majority in all genomes (about 90% in Archaea, about 93% in Eubacteria, 89% in Saccharomyces cerevisiae and 90% in Caenorhabditis elegans). Aeropyrum pernix among Archaeae and Deinococcus radiodurans among Eubacteria have the lowest fraction of high complexity proteins (75% and 80% respectively). Further, it was observed that a few bacterial pathogens (Mycobacterium tuberculosis, Pseudomonas aeruginosa) have high fraction of low complexity proteins. The program ScanCom is available from the authors as a PERL script (UNIX system).