Fig. 4

Flowchart to detect the motifs representative of high/low expression categories: To detect the motifs enriched in high expression class, the very low expression class was used as negative control. The reverse was done to search for motifs enriched in very low expression class. Of all the very highly expressed genes, two types of training sets were made, 2/3 and 1/2 of genes were used as training set and rest as test set. It was repeated in 3 rounds by randomly selecting genes for training and test set. The predicted motifs were checked for enrichment in test set and also in other expression classes (high, medium, and low)