Assessing tools for genome wide prediction of PurR transcription factor binding sites in Escherichia coli

Abstract

Now a day’s DNA motif prediction tools considering weight matrices of regulatory elements rather than overrepresented patterns or consensus signature sequence to identify motifs in whole genome. The present work describes evalution of motif discovery tools based on weight matrices at genomic level. For this, we selected experimentally reported 11 PurR transcription factor regulatory promoters of E. coli as reference data and developed weight matrices through available tools e.g., AlignACE, MEME, D-Matrix, Consensus and than compared the prediction accuracy with available matrix at RegulonDB database. Genome wide PurR binding sites were first predicted on reference data and than on unknown data set of 951 known E. coli promoters through motif matching program Patser and Possumsearch. For statistical validation sensitivity, specificity, false positive ratio and prediction accuracy were calculated as threshold dependent parameters and area under curve plots as threshold independent parameter. We concluded that weight matrix made by Consensus tool showed higher accuracy (99%) and best fitted ROC area (98%). It was also observed that true positive predictions were depends on the optimum cut-off score and length of promoter sequences. ©IJIB, All rights reserved.