Thesis (M.Sc.)--Chulalongkorn University, 2003

DNA spectra plot can fairly indicate exon locations. The main goal of this work is to apply the Digital Signal Processing techniques to predict exon locations more accurately. A group of genes from Caenorhibtidis elegans was used to test the methods. In spectral analysis process, a DNA sequence was transformed to binary sequences by 4 mapping rules depending on nucleotides, amino acids, hydrophobicity, and codon bias, respectively. The triplet-periodicity discovered in exons stands out only when we used 4 binary indicator sequences corresponding to 4 nucleotides. The optimized spectral content measure proposed by Anasstasiou was developed. Alternating samples used in the optimization problem shown that it is not necessary to use a very large sample. The attempt to discriminate between exons and introns of the same genes succeeded in reducing unwanted peaks.