Table 1.
Correlation between PWM scores and expression
| Organism | Experiment | Biological process/tissue | Motif | Average information content | %RIV | P-value |
|---|---|---|---|---|---|---|
| Yeast | Microarray | Cell cycle (G2) | Mcm1 | 0.59 | 4.9 | 3.4 × 10−9 |
| Human | Microarray | Liver | HNF-1 | 0.62 | 2.2 | 1.6 × 10−6 |
| Human | Microarray | Pancreas | C/EBP beta | 0.76 | 2.5 | 2.5 × 10−7 |
| Human | Microarray | Cell cycle (G1/S) | E2F | 1.74 | 5.9 | 1.7 × 10−12 |
| Mouse | Microarray | Liver | HNF-1 | 0.62 | 2.1 | 2.1 × 10−6 |
| Mouse | Microarray | Liver | HNF-4 | 0.60 | 2.0 | 4.1 × 10−6 |
| Human | ChIP-chip | Liver/HNF-1 | HNF-1 | 0.62 | 14.7 | 8.7 × 10−37 |
Comparison of percent reduction in variances (%RIVs) between yeast and mammals is shown. Single linear splines were used to obtain the %RIV. Yeast cell-cycle data were obtained from Spellman et al (1998), liver and pancreas microarray data from Su et al (2004), liver ChIP-chip data for HNF-1 from Odom et al (2004) and human cell-cycle data from Whitfield et al (2002). Mcm1 matrix was obtained from Bussemaker et al (2001); all other matrices were obtained from TRANSFAC and JASPAR databases. P-values were calculated using an F-test (see Materials and methods). Average information content refers to the information content of the position weight matrix, averaged over all the positions (columns). It is 2.0 for non-degenerate words and 0 if all positions were N's (see Materials and methods).