摘要： 在对蛋白质预测结构进行聚类的过程中，常用的均方根偏差、TM-score、GDT-TS等相似性度量方法仅反映了结构之间的距离关系而未考虑结构之间的能量关系。针对上述问题，对候选结构进行距离度量，计算两两之间的能量差异，并以此设置权重，对相似性矩阵进行修改。通过在13个数据集上的实验表明，采用能量差异对相似性矩阵进行加权后的聚类结果优于加权之前。
关键词:
蛋白质结构预测,
聚类算法,
均方根偏差,
相似性矩阵
Abstract: The structure metrics such as Root Mean Square Deviation(RMSD), TM-score, and GDT-TS etc. reflect only distance instead of energies difference between any decoys. This paper calculates difference of energies for any pair of decoys, modifies the similarity matrix with the differences, and clusters decoys with these similarity matrices respectively. Experimental results on 13 decoys sets show that the strategies of weighting similarity based on energies for clustering protein structures can improve the performance of clustering.
Key words:
protein structure prediction,
clustering algorithm,
Root Mean Square Deviation(RMSD),
similarity matrix
中图分类号:
黄旭, 吕强, 吴进珍, 钱培德. 基于能量的蛋白质结构聚类距离加权策略[J]. 计算机工程, 2010, 36(21): 173-174,177.
HUANG Xu, LV Jiang, TUN Jin-Zhen, JIAN Pei-De. Strategies of Weighting Distance for Clustering Protein Structures Based on Energy[J]. Computer Engineering, 2010, 36(21): 173-174,177.