基于周期3临近特征的外显子预测方法

A Near Period-3 Feature Strategy to Improve the Extron Prediction

为提高基因编码序列(外显子)的预测精度,提出了一种周期3临近特征外显子预测方法。分别提取外显子和内含子的周期3临近功率谱聚类特征值作为模板特征,将未知外显子和内含子区间的DNA序列分帧移动,依此提取每帧特征并与模板特征实现不同权重的欧氏距离的识别预测。通过改变基因序列不同特征值、特征值数、帧长、权重,将该算法与周期3功率谱特征预测法比较,实验结果表明,特定参数条件下该算法的预测精度优于周期3预测法。

To improve the accuracy of the gene encoding （extron） prediction, near period-3 feature extrons prediction algorithm is proposed. Near period-3 clusting power spectrum of extrons and introns are extracted as template feature, DNA sequence whose extrons and introns ranges are unknown is divided into frames and moved. Compared with the template feature, the prediction of the Euclidean distance with different weights is realized from each frame. By changing the different feature, number, frame length, gene sequence weight and comparing with period-3 algorithm, the experiment results show that the prediction accuracy of the proposed algorithm is better than that period-3 algorithm.