Examining Start Sites of Predicted Bacterial Genes Through Protein Homology Search

Abstract:

Due to high density and relatively simple structure,genes in bacterial genomes can be predicted with high accuracy.An important remaining issue in bacterial gene prediction is to precisely identify the gene start sites.At present,the most widely used automatic genome annotation pipelines result in the disagreed prediction of start sites for about 1/5 of the total genes.There is no effective method available to improve the start site prediction by using statistical characteristics of gene structure.This manuscript attempts to find new clues for the deduced start sites through protein homology search.Our analysis of the genome of Gluconobacter oxydans 621H shows that this method can substantially improve the start site prediction.