基于小波特征提取和支持向量机的蛋白质二级结构预测

Protein Secondary Structure Prediction Based on Wavelet Feature Extraction and Support Vector Machine

摘要: 蛋白质的结构对理解蛋白质的生物学功能意义重大，蛋白质结构的预测就能预测和理解未知蛋白质生物学功能的作用，并且蛋白质二级结构的预测是对蛋白质结构的预测起决定性作用的，在蛋白质二级结构预测的研究中，将蛋白质单个残基用位置特异性打分矩阵(PSSM)进行编码，取窗口后可以将一个蛋白质残基表示成一个2维的伪图像平面，在原位置特异性打分矩阵数据平面的基础上，用小波变换提取到伪图像平面不同分辨率水平上的低频特征和高频特征与原PSSM平面数据当作一个蛋白质残基携带的样本信息，并用支持向量机对预测进行训练模型。

Abstract: The structure of proteins is very important for understanding the biological function of proteins. The prediction of protein structure can predict and understand the function of biological functions of unknown proteins; however, the prediction of protein secondary structure plays a decisive role in the prediction of protein structure. In the study of protein secondary structure prediction, a single residue of a protein is encoded by position-specific-score-matrix (PSSM). After a data window is taken, a protein residue can be represented as a 2-dimensional pseudo-image plane, thus could further use the wavelet method to extract multi-resolution based features both on high frequency and low frequency from original pseudo-image, these extracted wavelet-based features with the PSSM matrix together can be taken as sample information carried by a protein residue, and the training model used is support vector machine.