Identifying Temporal Gene Networks Using Signal Processing Metrics on Time-Series Gene Expression Data

Abstract

A gene network refers to the knowledge of the activators and inhibitors of all genes. The genes themselves are believed to function as regulators of other genes. Most work done so far either ignores time delay in gene regulation or assumes that it is constant. We here propose the use of signal processing metrics like correlation techniques to find the gene interactions. Also, a post-processing stage is developed to remove false interactions among genes due to common parents, and dynamic correlation thresholds are used for selecting suitable correlation coefficients for constructing the gene network. The proposed correlation based network learning algorithm (CBNL Algorithm) considers the multi time delay relationships among the genes, and therefore estimates the temporal gene network. The implementation of our method is done in MATLAB and experimental results on Saccharomyces cerevisiae expression data and comparison with other methods indicate the effectiveness of the method