a model-based method for gene dependency measurement基于模型的方法,基因依赖测量.pdfVIP

A Model-Based Method for Gene Dependency Measurement 1 2 1 1 1 1 Qing Zhang , Xiaodan Fan , Yejun Wang , Mingan Sun , Samuel S. M. Sun , Dianjing Guo * 1 School of Life Sciences and the State Key Laboratory of Agrobiotechnology, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China, 2 Department of Statistics, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China Abstract Many computational methods have been widely used to identify transcription regulatory interactions based on gene expression profiles. The selection of dependency measure is very important for successful regulatory network inference. In this paper, we develop a new method–DBoMM (Difference in BIC of Mixture Models)–for estimating dependency of gene by fitting the gene expression profiles into mixture Gaussian models. We show that DBoMM out-performs 4 other existing methods, including Kendall’s tau correlation (TAU), Pearson Correlation (COR), Euclidean distance (EUC) and Mutual information (MI) using Escherichia coli, Saccharomyces cerevisiae, Drosophila melanogaster, Arabidopsis thaliana data and synthetic data. DBoMM can also identify condition-dependent regulatory interactions and is robust to noisy data. Of the 741 Escherichia coli regulatory interactions inferred by DBoMM at a 60% true positive rate, 65 are previously known interactions and 676 are novel predictions. To validate the new prediction, the promoter sequences of target genes regulated by the same transcription factors were analyzed and significant motifs were identified. Citation: Zhang Q, Fan X, Wang Y, Sun M, Sun SSM, et al. (2012) A Model-Based Method for Gene Dependency Measurement. PLoS ONE 7(7