Figure 1

Conceptual pipeline for TF→target predictions based on correlation and/or CRE analysis: Performed on Regulated Genes. Each data set (3 microarray datasets covering 4 transcription factors, see Table 1) has been subjected to the analysis outlined in Figure 1. For each transcription factor, we predicted 8 different gene lists of potential target genes* using Correlation analysis, CRE detection methods, or the combination of Correlation and CRE analysis methods. The first four lists were predictions made using only Correlation or CRE detection methods. The succeeding four gene lists, were generated by the intersection of predicted targets from Correlation and CRE methods. Misregulation was determined by comparing wild type microarray data, in which the TF is shown to be regulated, and micoarray data from TF mutant/transgenic (transgenic over-expressor). Both data sets contain replicates (at least 2 replicates) from each dataset. Performance was then evaluated using specificity, sensitivity NPV, PPV and the F-measure as statisitcal measures. TFs: Transcription Factors, OVER-CRE: over-represented CRE binding sites, >= 1: greater than equal to 1 CRE binding site

*Note that the term potential target gene refers to genes that are predicted to be misregulated in the transgenic plants. The analysis does not make any assumption whether this interaction is direct or indirect, or that these predictions are at all inclusive. The possibility that this interaction could be direct is addressed further in the Discussion.