Compromise of Multiple Time-Resolved Transcriptomics Experiments Identifies Tightly Regulated Functions

Abstract

With the advent of high-throughput technologies for data acquisition from different components (i.e., genes, proteins, and metabolites) of a given biological system, generation of hypotheses, and biological interpretations based on multivariate data sets become increasingly important. These technologies allow for simultaneous gathering of data from the same biological components under different perturbations, including genotypic variation and/or changes in conditions, resulting in so-called multiple data tables. Moreover, these data tables are obtained over a well-chosen time domain to capture the dynamics of the response of the biological system to the perturbation. The computational problem we address in this study is twofold: (1) derive a single data table, referred to as a compromise, which captures information common to the investigated set of multiple tables and (2) identify biological components which contribute most to the determined compromise. Here we argue that recent extensions to principle component analysis called STATIS and dual-STATIS can be used to determine the compromise on which classical techniques for data analysis, such as clustering and term over-enrichment, can be subsequently applied. In addition, we illustrate that STATIS and dual-STATIS facilitate interpretations of a publically available transcriptomics data set capturing the time-resolved response of Arabidopsis thaliana to changing light and/or temperature conditions. We demonstrate that STATIS and dual-STATIS can be used not only to identify the components of a biological system whose behavior is similarly affected due to the perturbation (e.g., in time or condition), but also to specify the extent to which each dimension of the data tables reflect the perturbation. These findings ultimately provide insights in the components and pathways which could be under tight control in plant systems.