Figure 1

Schema of our method. We first calculate correlations between phenotypes and expression values as meta-expression values, while preparing a sequence feature table by searching promoter sequences for cis-regulatory motifs. Cis-regulatory motif data are prepared from two different sources: already known motifs, which are downloaded from databases, and de novo identified motifs, which were discovered by an ab initio motif finder program, DME. Then, associations between sequence features and meta-expression values were inferred by structure learning of Bayesian networks.