Fig. 2.

Predicting cold-responsive genes in maize. (A and B) Baseline expression control. (A) Ability to predict whether a gene will be differentially expressed in response to cold based solely on baseline expression level before and after controlling for variation in gene-expression level. (B) Distribution of average FPKM values of cold-responsive genes (CR) and nonresponsive genes (NR), and training sets resampled from genes in 12 bins with balanced gene-expression levels (darker color). (C) ROC curves showing the performance of different maize models trained to predict cold-responsive gene expression using different types of features to describe genes. (D) PR curves showing the performance of different maize models trained to predict cold-responsive gene expression using different types of features to describe genes.