Gene Prediction by Multiple Spliced Alignment

Abstract

With recent advances in sequencing technologies, a huge amount of DNA sequences become available year after year. In order to obtain useful information on these sequences, we need to process them in search of biologically meaningful regions. The genes are amongst the most important regions of a genome and the task of locating them in a DNA of interest is called the gene prediction problem. This problem can be addressed in several ways, and one of the most promising methods relies on homology information between the genomic DNA and previous annotated sequences (proteins, cDNAs and ESTs). In this paper we generalize a traditional formulation of the gene prediction problem and use this new formulation in the development of three gene identification tools. All these programs were tested on a benchmark of 240 human genomic sequences and the results obtained compare favorably with those achieved by other gene prediction tools available in the literature.