Figure 1.
Identification of Bidirectional Promoters
The diagram depicts how bidirectional promoters were identified from the UCSC Human Genome Browser annotations, including Known Genes, mRNA, and spliced ESTs. The number of previously identified bidirectional promoters for each category is marked with an asterisk (*) for comparison. Promoters identified from more than one source were counted only once. Priority was given to promoters identified from Known Genes followed by GenBank mRNA, and then spliced ESTs.
Table 1.
Somatic Ovarian Cancer Genes Regulated by Bidirectional Promoters
Figure 2.
Clustering Analysis To Assess Bidirectional Promoters
Clustering methods include microarray expression data (red), GO categories (blue), and psi-blastP alignments (green). The plots show the ratio of bidirectional promoters calculated for the cumulative number of genes found at each promoter distance. The named breast or ovarian cancer gene served as the reference gene for each type of cluster. The computational approach used to count the number of bidirectional promoters within each cluster is described in Methods. GO annotations (blue) were omitted when no data was available.
Figure 3.
Frequency of Bidirectional Promoters
Each of 16,078 genes represented in the microarray data was grouped with its 500 most-related genes into an expression cohort. The ratio of bidirectional promoters in each expression cohort was plotted for three categories: (A) all genes, (B) DNA-repair genes, or (C) brain-specific genes.
Figure 4.
Co-Regulation of Bidirectional Promoters
(A) The number of bidirectional promoters is graphed for the breast and ovarian cancer genes (solid brown line) and their bidirectional partner genes (solid black line). All graphs have the same axes: (x) the ranking of genes in the expression cluster according to their distance from the target gene and (y) the number of bidirectional promoters identified at this distance.
(B) The correlation coefficient between each breast and ovarian cancer gene and its partner was calculated using the expression data in all 79 tissues. The correlation coefficients are plotted from the largest to smallest values; therefore, the tissue identity may change between plots to accommodate the ordering scheme.
Figure 5.
Conserved ETS Factor Binding Sites
(A) Transcription factor-binding sites are shown for the bidirectional promoters of ERBB2, FANCD2, and BRCA2. The vertical arrow points to the binding site preferred by the ETS factor family member ELK1. The legend indicates the identity of all binding sites.
(B) The sequence level view of these ETS factor binding sites are shown as snapshots from the UCSC Human Genome Browser. Horizontal arrows indicate the direction of transcription. Position 0 in (A) corresponds to the transcription start site of the cancer genes in (B). The annotation track labeled “Vertebrate Cons” represents the multiple-sequence alignment and conservation track for the eight-way vertebrate genome comparison at the UCSC Human Genome Browser. The putative ELK1-recognition sequences are boxed. The partner to BRCA2 is not within the range shown.