Email updates

Keep up to date with the latest news and content from BMC Genomics and BioMed Central.

Open Access Highly Accessed Methodology article

On detection and assessment of statistical significance of Genomic Islands

Raghunath Chatterjee1, Keya Chaudhuri1 and Probal Chaudhuri2*

Author Affiliations

1 Molecular & Human Genetics Division, Indian Institute of Chemical Biology, Jadavpur, Kolkata – 700 032, India

2 Theoretical Statistics and Mathematics Unit, Indian Statistical Institute, 203, B.T. Road, Kolkata – 700 108, India

For all author emails, please log on.

BMC Genomics 2008, 9:150  doi:10.1186/1471-2164-9-150

Published: 1 April 2008

Abstract

Background

Many of the available methods for detecting Genomic Islands (GIs) in prokaryotic genomes use markers such as transposons, proximal tRNAs, flanking repeats etc., or they use other supervised techniques requiring training datasets. Most of these methods are primarily based on the biases in GC content or codon and amino acid usage of the islands. However, these methods either do not use any formal statistical test of significance or use statistical tests for which the critical values and the P-values are not adequately justified. We propose a method, which is unsupervised in nature and uses Monte-Carlo statistical tests based on randomly selected segments of a chromosome. Such tests are supported by precise statistical distribution theory, and consequently, the resulting P-values are quite reliable for making the decision.

Results

Our algorithm (named Design-Island, an acronym for Detection of Statistically Significant Genomic Island) runs in two phases. Some 'putative GIs' are identified in the first phase, and those are refined into smaller segments containing horizontally acquired genes in the refinement phase. This method is applied to Salmonella typhi CT18 genome leading to the discovery of several new pathogenicity, antibiotic resistance and metabolic islands that were missed by earlier methods. Many of these islands contain mobile genetic elements like phage-mediated genes, transposons, integrase and IS elements confirming their horizontal acquirement.

Conclusion

The proposed method is based on statistical tests supported by precise distribution theory and reliable P-values along with a technique for visualizing statistically significant islands. The performance of our method is better than many other well known methods in terms of their sensitivity and accuracy, and in terms of specificity, it is comparable to other methods.