基于离散时序基因表达数据的双聚类算法  

Bicluster algorithm on discrete time-series gene expression data

作　　者：许涛[1] 尚学群[1] 杨蜜静 王淼[1] 

机构地区：[1]西北工业大学计算机学院计算机软件与理论系,西安710129

出　　处：《计算机应用研究》2013年第12期3551-3556,3567,共7页Application Research of Computers

基　　金：国家"973"计划资助项目(2012CB316203);国家自然科学基金资助项目(61272121)

摘　　要：目前应用于基因表达数据上的双聚类算法大多是基于真实数据提出的,因此易受噪声干扰,且这些算法很少考虑样本间的时序性。提出了一种有效的时间点连续的双聚类挖掘算法DTCB,从离散的时序基因表达数据中挖掘出时间点连续的最大共表达双聚类。该算法使用了一种新的数据离散化方法,同时提出了三种在离散数据集下基因间的共表达关系;为了提高挖掘效率,DTCB使用了有效的剪枝和输出策略,可以在不产生候选集的情况下一次性挖掘出所有的最大共表达双聚类。通过实验分析,证明DTCB具有高效的性能和良好的鲁棒性,且结果具有较好的统计和生物意义。At present, the bicluster algorithms applied to the gene expression data were mostly based on real data. Therefore,they were susceptible to noise interference, and these algorithms rarely considered the time sequence between samples. This paper proposed an efficient time-continuous bicluster algorithm DTCB to mine the maximal time-continuous biclusters from the discrete time-series gene expression data. It used a new discretization method on gene expression data and defined three co-expression relations between genes in the discrete dataset. DTCB adopted several pruning and output techniques to improve the efficiency. It could produce maximal co-expression biclusters without candidate maintenance. The experimental results show that DTCB has efficient performance and better robustness. Simultaneously,the results can be of more statistical and biological significance.

关 键 词：时序基因表达数据 双聚类 共表达 时间点连续 离散化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP301.6[自动化与计算机技术—计算机科学与技术]