Download the current version here or browse the Download directory for other versions.

We focus on the case of genes from a taxonomically distant source. Our approach to the problem is the use of a Hidden Markov Model (HMM) that incorporates data gained from comparative genomes considered as putative sources into a model describing the host genome, thus detecting possible genomic islands. It uses synonymous Codon Usage as the statistical feature, characterising the origin of a gene, that is incorporated into the model's observation probabilities, and serving as a similarity measure. In a preliminary step, Codon Usage data from a large number of species has been merged into clusters, each belonging to one taxonomic group that serves as a putative source. In one run of the algorithm, each gene of a species under consideration is classified according to the prepared data.

Our program SIGI-HMM is a publicly available and ready-to-use tool implementing this model, showing good prediction capabilities in the sense that it was able to reproduce experimental results for a number of well-studied genomes. We regard the usability of SIGI-HMM as its main advantage compared to other approaches that require a large amount of preprocessing done by the user. In contrast to that, SIGI-HMM works fully automated and can be adjusted with a small number of parameters.