Fig. 6
De novo secondary structure motif prediction. a Scheme for the de novo secondary motif prediction. The search was done over the sets of bins generated as in Fig. 5a, Step 1 and 2. Pairwise alignment considering secondary structure was performed on the sets of bins, and conserved and highly scored alignments were selected (Step 3). Using the alignments as “seeds”, SCFG model was generated (Step 4). Motifs were searched in all the 3′ UTRs in the clusters (Step 5-1), then hits were aligned (Step 5-2) and new SCFG model was generated (Step 5-3). This cycle was performed until no new hits were obtained in the searching step. b The resulting over-represented motifs in Cluster I. Motifs with q < 1 × 10−4 and 10 or more hits in the cluster were shown. Motifs in red letters, those in blue letters, or those in black letters represent similar motifs, respectively. c Representative structures of motif.0.271 and the top 3 of preferentially associated GO terms in genes having the motif. d Representative structure of motif.0.170 and the top 3 of preferentially associated GO terms in genes having the motif