Fig 2. Bivariate ETM for flowering time in A. thaliana, for different prediction methods (solid lines) and univariate analysis using observed flowering time, the most correlated environmental variable and the different trait predictions (dotted lines).

Left panel: number of known flowering genes recovered, as a function of total number of genes considered. Right panel: corresponding enrichment probabilities (-log10(p)). The following methods were used: CCA (purple), LM (green), RF (brown), EN (blue), analysis using summer day length and observed flowering time are marked in black and red respectively. Enrichment is defined as the probability of recovering k out of m genes by chance, under the hypergeometric distribution. The area in gray marks the 5% upper and lower percentiles based on 200 permutations of the univariate/bivariate traits.