Investigations of Escherichia coli promoter sequences with artificial neural networks: new signals discovered upstream of the transcriptional startpoint

We present a novel method for using the learning ability of a neural network as a measure of information in local regions of input data. Using the method to analyze Escherichia coli promoters, we discover all previously described signals, and furthermore find new signals that are regularly spaced along the promoter region. The spacing of all signals correspond to the helical periodicity of DNA, meaning that the signals are all present on the same face of the DNA helix in the promoter region. This is consistent with a model where the RNA polymerase contacts the promoter on one side of the DNA, and suggests that the regions important for promoter recognition may include more positions on the DNA than usually assumed. We furthermore analyze the E. coli promoters by calculating the Kullback Leibler distance, and by constructing sequence logos.