Finding pathogenicity islands in genome data with Multiscale Neural-Network

A novel technique for finding pathogenicity islands in genome data with multiscale neural network (MNN) is present. Firstly, decomposes different scales of genomic sequences signal into different scales of wavelets (local frequencies), and shrinking the coefficients of each scale of wavelets by means of a separate multilayer perception neural network, reconstruct the result sequence signal from the smoothed coefficients by applying the Inverse Wavelet Transform. Secondly,detect G+C patterns in genomes by comparing the result sequence with original sequences. The results on G+C patterns analysis of Dradiodurans chromosome I and N.serogroup A strain Z2491 are presented. The findings show that MNN is a powerful tool to detect pathogenicity islands in genome data.