Pentamer vocabularies characterizing introns and intron-like intergenic tracts from Caenorhabditis elegans and Drosophila melanogaster.

Overall compositional properties at the level of bases, dinucleotides and longer oligos characterize genomes of different species. In Caenorhabditis elegans, using recurrence analysis, we recognized the existence of a long-range correlation in the oligonucleotide usage of introns and intergenic regions. Through correlation analysis, this is confirmed here to be a genome-wide property of C. elegans non-coding portions. We then investigate the possibility of extracting a typical vocabulary through statistical analysis of experimentally confirmed introns of sufficient length (>1 kb), deprived of known splice signals, the focus being on distributed lexical features rather than on localized motifs. Lexical preferences typical of introns could be exposed using principal component analysis of pentanucleotide frequency distributions, both in C. elegans and in Drosophila melanogaster. In either species, the introns' pentamer preferences are largely shared by intergenic tracts. The pentamer vocabularies extracted for the two species exhibit interesting symmetry properties and overlap in part. A more extensive investigation of the interspecies relationship at the level of oligonucleotide preferences in non-coding regions, not related by sequence similarity, might form the basis of new approaches for the study of the evolutionary behaviour of these regions.