一种提高外显子预测的改进周期3消噪策略

为提高外显子的预测精度，文中用到了一种改进的算法-PPA算法（ post-processing algorithms ）．PPA算法通过比较滤波前后DNA信号N/3处的频谱分量的方法来去除噪声．将PPA算法分别应用到处理离散傅立叶变换（DFT），成对的频谱分析（ PSC）和改进的Gabor小波变换（ MGWT）的周期3系数来实现周期3的去噪．算法的性能在HMR195，Burset/Gui-go570，Asp673个数据库中进行了评估，并把受体工作特征（ ROC）、特异性、灵敏度曲线3个方面作为评测标准．通过实验分析，验证了PPA算法能够有效地提高外显子的预测精度，同时基于小波变换的策略确保了模型的独立性．

An improved cycle-3 denoising strategy to improve the exon prediction

In order to improve the accuracy of the exon prediction , an improved algorithm -PPA algorithm is used in this paper .PPA compares the N/3 spectral components of DNA signal with the suppressed DNA signal . PPA algorithm were applied to process period -3 coefficients of the discrete Fourier transform （ DFT）, Gabor wavelet transform （MGWT） and paired spectrum analysis （PSC） to remove noise.Performance of the algorithm has been evaluated on HMR195, Burset/Guigo570, and Asp67 datasets, and the receptor operating characteris-tic （ROC）, specificity, sensitivity curve are used as evaluation standards .Through experimental analysis, we show that the PPA algorithm can effectively improve the prediction accuracy of the exon , and ensure the inde-pendence of the model based on wavelet transform strategy .