Figure 1.
Choosing Pearson correlation coefficient cutoff value.
A, the actual number of edges and all possible edges among the non-singleton nodes as a function of correlation coefficient cutoff values. B, network density at different correlation coefficient cutoff values.
Figure 2.
Degree distribution of the drought-responsive gene co-expression network.
A, the preliminary network without the maximum connection. B, the final drought-responsive gene network.
Figure 3.
Mapping the modules onto the drought-responsive gene co-expression network.
The nodes are color-coded by modules and gray nodes represent genes unassigned to a module. The over-represented GO terms are shown for each module. Each pie chart represents the proportion of up- (yellow color) and downregulated (blue color) genes in the corresponding module.
Figure 4.
Heat map derived from the drought-responsive gene co-expression network.
A, heat map of our network. Genes in the rows and columns have been ordered by a MCL algorithm. Each of the colored bars along the top horizontal and left vertical axes represents a gene module. Genes that do not belong to any module are colored gray. B, heat map of the WGCNA network based on power (β = 4). The genes are colored by module assignment in A.
Figure 5.
Tissue-specific expression of genes in the drought-responsive modules.
The average expression level for each gene in different tissue types was calculated based on the normalized Affymetrix array data.
Figure 6.
Heat map of the putative cis-regulatory motifs enriched in the drought-responsive modules.
The motifs in the rows and columns have been ordered by simple hierarchical clustering. A gradient of colors represent the Pearson correlation coefficients between motifs. The black line rectangles in the heat map indicate the similar motifs with a correlation coefficient >0.65. Functions of the merged motifs are indicated in the right-sided table.
Table 1.
Module significance in response to drought stress in different rice varieties.
Figure 7.
Co-expression relationship of HSF with its related genes.
A, gene co-expression network and putative binding sites for HSF. B, the average expression levels of HSF and its neighborhood genes under PEG and ABA treatments. The error bars represent standard deviation from the mean.
Table 2.
Functional enrichment of drought-responsive hub genes.
Figure 8.
Distribution of hub gene clusters and the overlapped drought tolerance QTLs in rice chromosomes.
Red and blue filled rectangles represent hub gene clusters and QTLs for rice drought tolerance, respectively.