Outlier Filtering for Identification of Gene Regulations in Microarray Time-Series Data

Microarray technology provides an opportunity for scientists to analyze thousands of gene expression profiles simultaneously. Time-series microarray data are gene expression values generated from microarray experiments within certain time intervals. Scientists can infer gene regulations in a biological system by judging whether two genes present similar gene expression values in microarray time-series data. Recently, a great many methods are widely applied on microarray time-series data to find out the similarity and the correlation degree among genes. Existing approaches including traditional Pearson coefficient correlation, Bayesian networks, clustering analysis, classification methods, and correlation analysis have individual disadvantages such as high computational complexity or they may be unsuitable for some microarray data. Traditional Pearson correlation coefficient is a numeric measuring method which gives novel effectiveness on two sets of numeric data. However, it is not suitable to be applied on microarray time-series data because of the existence of outliers among gene expression values. This paper presents a novel method of applying Pearson correlation coefficient along with an outlier filtering procedure on the widely-used microarray time-series datasets. Results show that the proposed method produces a better outcome compared with traditional Pearson correlation coefficient on the same dataset. Results show that the proposed method not only can find out certain more known regulatory gene pairs, but also keeps rational computational time.