G蛋白偶联受体及其类型的预测

G蛋白偶联受体是非常重要的信号分子受体，其功能失调会导致许多疾病的产生。在前期工作的基础上，作者将序列特征分析与支持向量机技术结合起来，通过分析序列的特征差异，对G蛋白偶联受体分子及其类型进行识别。首次提取了G蛋白偶联受体对应的mRNA序列的绝对密码子使用频率作为特征，这主要因为它既包含了基因密码子使用偏性的信息，也包含了基因所编码蛋白的氨基酸组成信息。结果显示：在G蛋白偶联受体序列及其类型预测的问题中，设计支持向量机分类器时，最好选择使用包含基因序列绝对密码子使用频率和蛋白序列双联氨基酸使用频率两部分信息的组合特征作为特征，同时采用径向基核作为核函数。

Prediction of G-Protein Coupled Receptors and Their Type

G-protein coupled receptor is a very important signal molecule receptor and its dysfunction may lead to the emergence of many diseases. According to the previous studies, a method combining the feature analysis methods of sequences with support vector machine (SVM) technology was proposed for identifying GPCRs and their type by analyzing the characteristics of sequence differences. Especially, codon use frequencies of mRNA genes translating into GPCR proteins were first selected as the sequence feature, in respect that it is the inherently the fusion of both codon usage bias and amino acid composition signals. The results showed that the optimal SVM classifiers for predicting GPCR sequences and their type were designed by choosing the hybrid feature by combining codon use frequencies of mRNA genes and double amino acid use frequencies and using the RBF kernel as kernel function after considering the performance of all types of SVM classifiers.