Table 2.
Comparison of IIDB TFBS predictions with ChIP-chip data. Data are presented based on the genome annotations available from both NCBI and ENSEMBL. Note that the annotations differ in the number of predicted genes.
| Using NCBI coordinates | Using ENSEMBL coordinates | |||||||
| Promoter region mapped † | 1000 | 2000 | 1000 | 2000 | 1000 | 2000 | 1000 | 2000 |
| Number of genes | 1151 | 1151 | 1151 | 1151 | 1935 | 1935 | 1935 | 1935 |
| Unique ATF3 ChIP-chip hits | 978 | 1761 | 978 | 1761 | 1494 | 2750 | 1494 | 2750 |
| Conserved promoter regions containing ATF3 TFBS ◇ | 833 | 979 | 833 | 979 | 1329 | 1550 | 1329 | 1550 |
| Percentile threshold ■ | 0.05 | 0.05 | 0.01 | 0.01 | 0.05 | 0.05 | 0.01 | 0.01 |
| ATF3-group matrices hits* | 792 | 1187 | 212 | 299 | 1333 | 2029 | 337 | 474 |
| ATF3-group matrices within a ChIP-chip segment | 442 | 664 | 110 | 165 | 710 | 1031 | 196 | 272 |
| %overlap between ChIP-chip data & predictions ◉ | 55.8 | 55.9 | 51.8 | 55.2 | 53.2 | 50.8 | 58.1 | 57.4 |
Notes: ◇ Conserved regions were mapped from the human data of Xie et.al [14]. ■ Threshold refers to the p-value below which predicted TFBS are considered significant. † Numbers refer to length of promoter annotated, in base pairs upstream of the transcription start site. * Since ATF3 binding sites have a strong overlap with CREB binding sites, we used a combined PWM including three ATF matrices and nine CREB matrices to calculate the ATF3 hits. Overlapping hits were collapsed into one as described in the main text. ◉ ChIP fragments in these experiments were estimated to have an average length of approximately 500 bp to 1 Kbp. To determine the coordinates of a ChIP-chip hit, we estimated the center of gravity of a bound region using a moving average filter, then set the start/end coordinates to be +/- 300 bp.