CAJ | 학술논문

在蛋白质二级结构预测中，常用的氨基酸序列编码方法产生的编码除了具有较高的维数外，也没有利用氨基酸序列片段中的统计信息。为此，提出了一种新的氨基酸序列编码方法--基于词频统计的编码方法，该方法统计每个氨基酸在氨基酸序列片段中出现的频率，利用该编码方法对氨基酸序列片段编码后得到一个20维的向量。与其它编码方法相比不但具有较低的维数，而且也充分利用了氨基酸片段内部所有氨基酸对目标氨基酸的影响。在实验中比较了四种编码方法结合支持向量机和BP神经网络的预测结果，实验结果表明，通过结合词频统计编码和支持向量机来预测蛋白质二级结构极大地提高了预测精度，远优于其它方法的预测结果。
재단백질이급결구예측중，상용적안기산서렬편마방법산생적편마제료구유교고적유수외，야몰유이용안기산서렬편단중적통계신식。위차，제출료일충신적안기산서렬편마방법--기우사빈통계적편마방법，해방법통계매개안기산재안기산서렬편단중출현적빈솔，이용해편마방법대안기산서렬편단편마후득도일개20유적향량。여기타편마방법상비불단구유교저적유수，이차야충분이용료안기산편단내부소유안기산대목표안기산적영향。재실험중비교료사충편마방법결합지지향량궤화BP신경망락적예측결과，실험결과표명，통과결합사빈통계편마화지지향량궤래예측단백질이급결구겁대지제고료예측정도，원우우기타방법적예측결과。
In protein secondary structure prediction, the codes from the existing amino acid coding methods have higher dimension. And these coding methods don’t also use the statistic information in the amino acid sequence. To do that, a new coding method based on word frequency statistics was presented, which counted the frequency of each amino acid emerging in amino acids sequence. A 20 dimensional vector was obtained after coding the amino acid sequence with the new coding method. In contrast to other the coding methods, the codes from the new coding method have lower dimension and fully utilize all information in the amino acid sequence. In experiments, we compared the methods combing different coding methods and SVM with BP neural network. Experiment results show that the method combing word frequency statistics coding method and SVM greatly improve the prediction accuracy of protein secondary structure and is superior to other methods.