| ECCB 2002 Poster sorted by: Author | Number Next | Previous poster (in order of the view you have selected) |
Title: Interpreting Patterns of Codon Usage in Bacterial Genomes | P97 |
Mahony, Shaun; Golden, Aaron; McInerney, James shaun.mahony@nuigalway.ie National Centre for Biomedical Engineering Science, NUI Galway |
| Most codons (with the exception of Met and Trp) have at least one synonymous alternative. The choice of synonymous codons is not random and codon usage patterns vary interspecifically (between genomes) and intraspecifically (between genes in the same genome). In this work, we develop a variation of Kohonen's Self-Organizing Map neural network architecture[1] that has the ability to identify common or repeated patterns of codon usage within a dataset of genomic sequences. It is then possible to show the likelihood that a gene is using a typical codon usage pattern as opposed to a relatively random one. We can also explore the possibility that certain combinations of codons are used by different categories of genes. In the past, it has been noted that in some organisms there are differences in synonymous codon usage between highly-expressed and lowly-expressed genes[2]. We have used the Self-Organizing Map approach to investigate the existence of more subtle patterns. As an example, we have applied our method to the analysis of the Neisseria meningitidis (serogroup B) genome published by The Institute for Genomic Research[3] (TIGR). TIGR has also published a listing of 2158 protein-coding regions, some of which are known to be real and have been assigned a biological role and some of which are simply predictions and as such are "hypothetical''. Unsurprisingly, we have found that a significant proportion of the "hypothetical proteins" display codon usage patterns that are significantly different to the codon usage patterns of known genes, while some genes in this category display markedly similar patterns to known genes. A smaller proportion of the known genes also display random codon usage patterns. Possible reasons for this variation are explored. One of the more interesting observations is that there is a strong correlation between particular codon usage patterns found in the known genes and the function of those genes. This is the first suggestion that gene function might play a role in synonymous codon selection. This suggests the possibility that this method may also be used to make inferences about gene function. |
| [1] Kohonen, T. (1995) Self-Organizing Maps [2] Ikemura, T. (1981) J. Mol. Biol.151, 389-409 [3] Tettlin, H. et al. (2000) Science 287, 1809-1815 |