Neural network system for prediction of RNA polymerase II promoters

摘要

One of the most difficult problems in the analysis of eucaryotic genes is the detection of RNA polymerase II promoter regions. Although promoter regions vary in the primary DNA sequence, a basic group of core promoter elements has been suggested in the literature. Many human promoter sequences contain a TATAA sequence element at approximately 30 bases upstream of the cap site (transcription start site). Other elements are the GC box which binds SPA and upregulates transcription, the CAAT box, and the ATG initiator codon. To characterize promoters, we constructed frequency matrices for each element using experimentally mapped human promoter regions. Additionally, we constructed histograms for the distances separating the various elements. We then used a neural network to combine these informational elements. The output of the neural network is then processed using a set of expert rules which depend on GRAIL's ability to find exons in anonymous DNA. This improves the selectivity of promoter detection and reduces the false positive rate.