Link-based quantitative methods to identify differentially coexpressed genes and gene pairs

Abstract

Background: Differential coexpression analysis (DCEA) is increasingly used for investigating the global transcriptional mechanisms underlying phenotypic changes. Current DCEA methods mostly adopt a gene connectivity-based strategy to estimate differential coexpression, which is characterized by comparing the numbers of gene neighbors in different coexpression networks. Although it simplifies the calculation, this strategy mixes up the identities of different coexpression neighbors of a gene, and fails to differentiate significant differential coexpression changes from those trivial ones. Especially, the correlation-reversal is easily missed although it probably indicates remarkable biological significance.

Results: We developed two link-based quantitative methods, DCp and DCe, to identify differentially coexpressed genes and gene pairs (links). Bearing the uniqueness of exploiting the quantitative coexpression change of each gene pair in the coexpression networks, both methods proved to be superior to currently popular methods in simulation studies. Re-mining of a publicly available type 2 diabetes (T2D) expression dataset from the perspective of differential coexpression analysis led to additional discoveries than those from differential expression analysis.

Conclusions: This work pointed out the critical weakness of current popular DCEA methods, and proposed two link-based DCEA algorithms that will make contribution to the development of DCEA and help extend it to a broader spectrum.

Figure 1

A simplified illustration of the framework of five DCEA methods. A), two expression data matrices from two contrastive experimental conditions (X and Y) involving genes a, b, c, and d, visualized using shades from red (high expression level) to blue (low expression level), are transformed to a pair of coexpression networks. In the coexpression networks, gene pairs with absolute expression correlation values larger than 0.8 are connected with solid lines, while the rest with dashed lines. The line thickness is proportional to the absolute coexpression value. Red color highlights a negative coexpression value, and the grey-shaded node, gene a, is the one whose differential coexpression (dC) calculations are to be illustrated. B), different DCEA methods calculate the dC measure of gene a in different ways (see Results and Discussion for details).

Figure 2

Flowchart of the main steps involved in the two DCEA methods DCp and DCe. The upper and lower boxes comprise the major steps for DCp and DCe respectively, while outside the boxes are a few steps of the shared pre-processing process.

Figure 3

Limit fold change model applied to identify differentially coexpressed links (DCLs) from a simulated dataset pair in group C (dataset pair III). Each point represents a gene pair or a link characterized by its log correlation ratio and maximum absolute correlation value. A curve (red) y = a + (b/x) is used to fit the boundary outliers (black dots) determined by fraction δ, and points (blue) lying above the fitted curves are considered DCLs. A, same signed links; B, differently signed links.

Figure 4

Receiver-operating-characteristic (ROC) curves showing the capabilities of five DCEA methods in retrieving predefined DRGs. To simulate change of regulatory relationships, 10% links were removed (A), 10% switched (B), 5% removed and 5% switched (C) in a 1000-node network. Each curve is averaged over five simulations. The numbers in legend are areas under the ROC curves (mean and std).

Figure 5

Genes were organized into network modules via correlation-reversal relationships. A solid link connects a pair of genes with positive correlation in normal state but negative in disease state, while a dashed link connects genes with negative correlation in normal state but positive in disease state.