摘要: 利用氨基酸数字编码模型,将蛋白质序列转换为数字序列,根据偏序理论构建蛋白质哈斯矩阵。基于同一类型蛋白质哈斯矩阵图 具有相似图像纹理的假设,运用图像处理方法提取图像的几何矩作为伪氨基酸成分,对G-蛋白偶联受体类型分为2层进行预测,预测成功率分别为92.33%和85.48%。预测效果表明该方法是可行的。
关键词:
生物信息学,
G-蛋白偶联受体,
哈斯矩阵,
模糊K近邻算法,
Jackknife测试
Abstract: Amino acid numeric coding model is used to convert protein sequences into numeric sequences, and the protein Hasse matrix is constructed based on partial ordering. It is assumed that proteins belonging to a same class must have some sort of similar texture of the protein Hasse matrix images. Based on this, geometric invariant moment factors derived from the image are used as the pseudo amino acid components to predict G-Protein Coupled Receptor(GPCR) classes in two levels. Through a benchmark dataset, the overall success rates achieved by the test are 92.33% and 85.48% in the first and second level respectively. Experimental results show that this method is viable.
Key words:
bioinformatics,
G-Protein Coupled Receptor(GPCR),
Hasse matrix,
fuzzy K-nearest neighbor algorithm,
jackknife test
中图分类号:
肖绚, 徐培杰. 哈斯矩阵图的G-蛋白偶联受体类型预测[J]. 计算机工程, 2011, 37(18): 204-205.
XIAO Xuan, XU Pei-Jie-. G-Protein Coupled Receptor Classes Prediction of Hasse Matrix Image[J]. Computer Engineering, 2011, 37(18): 204-205.