Comparative Analysis of Genomic Island Prediction Tools

Genomic islands are segments of DNA in prokaryotic organisms that have characteristics that differ from other regions of the genome. The main characteristics are: GC content%; insertion sequences and direct repeats; genes associated with mobility, such as integrase and transferase; tRNAs that often flank these regions of DNA. The genomic composition of the genomic islands may present biological functions, in these cases, they are classified into: pathogenic island (PAI), metabolic island (MIs), island of resistance (IRs) and / or symbiosis island (SIs). Prediction tools for genomic islands use genomic comparison analysis and sequence composition analysis strategies. The comparative analysis seeks to identify distinct regions in sequences of nearby organisms, whereas the composition analysis evaluates and relates the composition of regions with the other regions of the genome. This research aims to evaluate qualitatively and quantitatively the predictors of genomic islands already developed, from sets of organisms. The tools were applied in 15 organisms, of which Escherichia coli CFT073 was chosen as a control for having genomic islands cured in vitro being considered our gold standard. The comparative results with the gold standard revealed that the GIPSy tools obtained the best performance, covering about 91% of the composition and region of the islands. Followed by Alien Hunter, 81%, IslandViewer4, 78%, Predict Bias, 31%, GI Hunter, 17% and Zisland Explorer with 16%. In the analysis of the intersection of the predicted regions, the tools Alien Hunter and Predict Bias showed similar results. IslandViewer4 and GIPSy achieved better results. The other tools had poor performance. The combination of the tools Alien Hunter, GIPSy and IslandViewer4 proved to be the best alternative in the prediction of the genomic islands in the organisms studied.