Network-based analysis of time series RNA-seq gene expression data by integrating the interactome and gene ontology information

Abstract

Monitoring the changes in gene expression patterns over time provides the distinct possibility of unraveling the mechanistic drivers characterizing cellular responses. Such time series gene expression data allow us to broadly "watch" the dynamics of the system. However, one challenge in the analysis of time series data is to establish and characterize the interplay between genes that are activated, deactivated or sustained in the context of a biological process or functional category. To address such challenges, novel algorithms are required to improve the interpretation of these data by integrating multi-source prior functional evidence. In this paper, we introduced a novel network-based approach to extract functional knowledge from time-dependent biological processes at a system level using time series mRNA deep sequencing data. First, a list of differentially expressed genes (DEGs) at each time point was identified. Second, GO terms that are enriched in each DEG list were identified. Third, the significance of interactions between DEGs in these GO terms at consecutive time points was measured. Finally, the significant interactions between DEGs in different GO terms were used to construct the interaction networks among GO terms between two consecutive time points, called GO networks. The proposed method was applied to investigate 1α, 25(OH)2D3altered mechanisms in zebrafish embryo development. GO networks were constructed over 4 consecutive time points. Results suggest that biological processes such as cartilage development and one-carbon compound metabolic process are temporally regulated by 1α,25(OH)2D3. Such discoveries could not have been identified with canonical gene set enrichment analyses. These results demonstrate that the proposed approach can provide insight on the molecular mechanisms taking place in vertebrate embryo development upon treatment with 1α,25(OH)2D3. Our approach enables the monitoring of biological processes that can serve as a basis for generating new testable hypotheses. Such network-based integration approach can be easily extended to any temporal-or condition-dependent genomic data analyses.