A Hybrid Method for the Recognition of Acceptor Splice Sites

Accurate prediction of exons in eukaryotic genes requires methods for exact acceptor and donor splice site prediction. In this paper, we report on a combination of a digital signal processing (DSP) technique and a data-driven statistical technique for the recognition of acceptor splice sites. The proposed hybrid method and the existing statistical techniques were compared on standard human datasets. Comparisons, (i) number of false positive (and specificity) at different sensitivity levels, and (ii) area under the receiver operating characteristic (ROC) curves, show that a DSP-statistical hybrid method outperforms the major existing techniques. By comparison with the acceptor splice site detection method used in GENSCAN, the proposed method reveals a consistent reduction in false positives, averaging 52.6%, and 56.1% reductions using the GENSCAN test and HS³D datasets respectively.