Improved gene prediction by resampling-based spectral analysis of DNA sequence

This paper proposes a new gene prediction method. In this approach, DNA sequences are resampled a number of times according to the codon bias hypothesis. In order to detect the period-3 pattern in the protein-coding regions (exons) of a gene, the resampled sequences are transformed to numerical signals through a proper mapping method, and spectral analysis is then applied to the resulting digital signals. Such resampling reduces spurious periodicities in the DNA sequence and thus achieves higher discriminative power in detecting exons from a DNA sequence. The advantage of this method is demonstrated by an experiment to the DNA sequence of the Stellate Cell Activation-associated Protein (STAP). The method can be an integrated part of more advanced gene prediction algorithms for improved accuracy.