基于Chernoff距离的GA-PLS法预测蛋白质二级结构研究

提出了用于预测蛋白质二级结构的Chernoff-GA-PLS算法。该方法首先是根据各个氨基酸残基的理化性质等自身所带的信息,计算出各样本到不同类别的Chernoff距离,进而根据Chernoff距离对蛋白质的氨基酸序列数据进行编码。最后由偏最小二乘进行蛋白质二级结构预测,并在整个算法过程中使用GA优化各个运行参数。为解决蛋白质二结构预测中的编码问题,提高预测结果的准确性和鲁棒性提供了一种新的思路。应用本方法对28个蛋白质共5789个氨基酸进行处理,获得的正确预测率达73.47%,研究结果表明,该方法预测结果明显高于目前运用单一方法获得的65%左右的预测准确率。由于该方法的预测误差小,易在Matlab上编程实现,计算过程中的参数意义明确和良好的可解释性,因此具有良好的应用前景。

Use GA-PLS method to predict protein secondary structures based on Chernoff distance

A novel method of Chernoff Distance-Genetic Algorithm-Partial Least Squares (Chernoff-GA-PLS) is built to predict the protein secondary structures. Based on the physical and chemical properties amino acids, the Chernoff distances from samples to groups H, E, C are calculated, and then the data of amino acids sequence are coded based on the Chernoff distance. Finally, protein secondary structures are predicted, and the GA is used to optimize parameters in the calculating course. The new thought of coding is advanced for predicting protein secondary structures and the veracity and robust of predicting results are improved. The correct rate of prediction is 73.47%. The study results indicate that the new method is able to get better results than those that use only a single method, whose correct prediction ratio is about 65%. The method is easily programmed in Matlab and the forecasting error is small. In general, the foreground of the new method applied is favorable.