Genomics and Computational Molecular Biology

this article permits us to mention only the most recent and major advances in techniques for gene identification. However, there are a number of other reviews and compendiums that cover this area in more depth [25, **26, 27, **28, 29, 30, 31, 32, 33]. In addition, Table 1 includes a list of Web pointers to most of the bioinformatics methods that are presented in this review. Computational methods for gene identification The first step in gene identification is the location of coding regions or open reading frames (ORFs). This task is simplified in bacteria due to the absence of splicing. Sequencing errors and translational frameshifting [34] can lead to partial protein sequences or interrupted open reading frames but these are often resolved during the early steps of gene identification by sequence similarity with proteins from other organisms [35, 36, 37, 38, 39, 40]. In the absence of homologous sequences in other organisms, and especially with short bacterial genes, probabilistic gene models (hidden Markov models) one can often identify biologically significant coding regions [41, 42]. Pairwise sequence homology Given a database of potential open reading frames, a large number of methods can be used to define the biological function of the putative proteins. The most commonly applied methods search for sequence similarity of the translated open reading frames with a database of known protein sequences [43, 44, **45, 46]. The search for gene function is usually carried out at the protein level to eliminate the redundancy of the genetic code. In addition, the use of amino acid substitution matrices that describe the acceptable replacements permits the discovery of even distantly related protein homologies [47, 48]. One of the most sensitive methods for comparing two ..