Gene expression
The PCA of gene expression log-fold changes revealed no obvious genome-wide parallelism or antagonism: just like with genetic divergence data, all principal components were similar in magnitude, and the shape of the gene cloud did not have any structure suggesting some predominant direction of overall gene expression change (Figure 5).
Figure 5. Principal component analysis of gene expression data (log-fold changes). Points are individual genes; vectors are original contrasts. Bar chart displays the proportion of variance explained by each principal component.
However, the gene set overlap analysis (Figure 6) revealed that all contrast-specific gene sets were significantly smaller than expected, while nearly all gene sets shared between three or more studies were significantly larger than expected. Interestingly, the two cases of adaptation to high CO2 did not show more overlap between each other than expected by chance, while sharing a higher-than-expected number of outlier genes with other datasets.
Figure 6. Gene expression gene set overlap analysis for the top 12.5% (a) and bottom 12.5% (b) quantiles of log-fold-change. Bar chart displays the number of outlier genes shared between contrasts identified by connected dots below; the first five bars are numbers of contrast-specific outlier genes. Asterisks above bars indicate gene sets that are significantly larger (red) or smaller (blue) than expected.