InterTransViewer: a comparative description of differential gene expression profiles from different experiments

Abstract

Meta-analysis of transcriptomic data from different experiments has become increasingly prevalent due to a significantly increasing number of genome-wide experiments investigating gene expression changes under various conditions. Such data integration provides greater accuracy in identifying candidate genes and allows testing new hypotheses, which could not be validated in individual studies. To increase the relevance of experiment integration, it is necessary to optimize the selection of experiments. In this paper, we propose a set of quantitative indicators for a comprehensive comparative description of transcriptomic data. These indicators can be easily visualized and interpreted. They include the number of differentially expressed genes (DEGs), the proportion of experiment-specific (unique) DEGs in each data set, the pairwise similarity of experiments in DEG composition and the homogeneity of DEG profiles. For automatic calculation and visualization of these indicators, we have developed the program InterTransViewer. We have used InterTransViewer to comparatively describe 23 auxin- and 16 ethylene- or 1-aminocyclopropane-1-carboxylic acid (ACC)-induced transcriptomes in Arabidopsis thaliana L. We have demonstrated that analysis of the characteristics of individual DEG profiles and their pairwise comparisons based on DEG composition allow the user to rank experiments in the context of each other, assess the tendency towards their integration or segregation, and generate hypotheses about the influence of non-target factors on the transcriptional response. As a result, InterTransViewer identifies potentially homogeneous groups of experiments. Subsequent estimation of the profile homogeneity within these groups using resampling and setting a significance threshold helps to decide whether these data are appropriate for meta-analysis. Overall, InterTransViewer makes it possible to efficiently select experiments for meta-analysis depending on its task and methods.