A new algorithm for predicting protein coding regions based on the hybird threshold

In protein coding regions prediction works, the predicted non-coding percentile (PNCP) is usually used as the threshold to separate the DNA sequences into two kinds of regions, i.e. protein coding regions and non-coding regions. In this paper, a new protein coding regions prediction algorithm based on the hybrid threshold is presented. First, the normalized power spectral density (PSD) of a DNA sequence is calculated using the prediction algorithm based on narrow pass-band filters (NPBF), and the maximum PSD value of the DNA sequence is used as the normalization standard in the NPBF algorithm. Second, the coding regions´ PSD curve is set up as a trapezoid model characterized by its slope and height. Third, the method for calculating the hybrid threshold which taking the slope and the height both into consideration is presented. Finally, the algorithm is performed on DNA dataset HMR195. Using the approximate correlation (AC) as the evaluation measure of prediction accuracy, the prediction results of the proposed algorithm reach to 0.48 for the dataset, which is much better than the modified Gabor wavelet transform algorithm. The prediction results are also presented in the form of q₉ proposed by Chun-Ting Zhang in 2002.