Boosting approach to exon detection in DNA sequences

A technique for increasing the discrimination between coding and non-coding regions of a DNA sequence during exon prediction is presented, based upon the boosting method employed in speech enhancement. The proposed method improves the likelihood of correctly identifying coding regions compared with existing genomic signal processing techniques, such as the DFT, IIR anti-notch filter, and multi-stage filters. Evaluations using receiver operating characteristic curves on 78 sequences from the HMR195 dataset show that the proposed method outperforms other methods by more than 3 to 13%.