Simple Methods of Finding Short Protein Coding Sequences

Eukaryotic genomes contain many conserved regions of unknown function. Accurately assessing the protein coding potential of these regions is a key step in annotation. We develop three protein coding measures that directly assess conserved regions in multiple sequence alignments of many species: one based on phase-shifts induced by alignment gaps, another based on the 3rd position mutation asymmetry in codons, and a third based on nucleotide composition asymmetry. The methods are easy to implement and require no training. Using a human-chimp-rat-mouse-chicken multiple alignment, these measures can classify coding regions as short as 30nt with greater specificity than single-genome measures using 120nt. Results from human-mouse and humanchicken alignments can be further improved by considering additional species; only the chimp genome proved uninformative. The phase-shift method is especially accurate. Contact: [email protected], [email protected]