基于模板匹配与塔式分解的蛋白质结构域分类

Structural classification of protein domain based on template match and pyramid decomposition

首先构造结构域的距离矩阵灰度图像;其次建立典型二级结构的距离函数,并分析所呈现的灰度模式;然后基于模板匹配和塔式分解,提出了结构域特征;最后在结构类和折叠子两个层次实施结构域分类。本方法在第一种验证策略的分类精度分别为90.7%和74.6%,使用第二种验证策略的为93.8%和78.7%。相比其他方法,具有更高分类精度和更低的特征维数,说明本方法更有效。

The classification of structural domain is one of important approaches which contribute to explore the mechanism of folding and the relationship of protein structure and its biological function.First,this paper mapped spatial structure of protein domain into Cα-Cα distance matrix which could be further regarded as gray texture image.Next,it modeled two distance functions for α helix and β strand/sheet by considering their geometrical properties,and used to find their gray patterns in distance matrix image respectively.After that,it applied the techniques of spatial template match and pyramid decomposition to present the composition feature of α helix and β strand and the multi-scale topology feature of β sheet respectively.Furthermore,in terms of the hierarchy of structural classification of proteins（SCOP）,performed domain classifications on structural class and fold levels respectively and compared with other methods.Finally,the results of domain classification show that the proposed method achieves the accuracies 90.7% and 74.6% in the first validation strategy,and 93.8% and 78.7% in the second validation strategy respectively.The comparison with other methods validates the presented method can perform domain classification effectively and outperform its competitors with both the higher classified accuracy and the more compacted dimension of feature vector.