Table 1.
Modeling c-Jun, GR and Hoxa2 binding using a logistic regression
| Evaluated parameters |
AP-1 motif |
GR motif |
Pbx motif |
||||||
|---|---|---|---|---|---|---|---|---|---|
|
(TGAC/GTCA)(11,926/271,973) |
(G-ACA---TGT-C)(1,393/27,176) |
(TGATTGAT)(638/59,802) |
|||||||
| PVE(%) | CV err. | AUC | PVE(%) | CV err. | AUC | PVE(%) | CV err. | AUC | |
|
A1. Peak |
10.3 |
0.04 |
0.76 |
2.9 |
0.05 |
0.63 |
0.5 |
0.01 |
0.57 |
|
A2. Background |
6.5 |
0.04 |
0.70 |
3.0 |
0.05 |
0.64 |
0.1 |
0.01 |
0.50 |
|
A3. Relative peak |
1.9 |
0.04 |
0.62 |
0.2 |
0.05 |
0.53 |
0.6 |
0.01 |
0.57 |
|
A. INOS (A1+A2+A3) |
11.5 |
0.04 |
0.77 |
3.9 |
0.05 |
0.66 |
0.6 |
0.01 |
0.57 |
|
B. Overlap with CGIs |
0.9 |
0.04 |
0.53 |
0.1 |
0.05 |
0.51 |
0.0 |
0.01 |
0.51 |
|
C. Cluster of canonical motifs |
5.5 |
0.04 |
0.66 |
0.6 |
0.05 |
0.51 |
0.1 |
0.01 |
0.51 |
|
D. Co-occurrence of 2nd motif* |
0.2 |
0.04 |
0.51 |
2.6 |
0.05 |
0.57 |
0.1 |
0.01 |
0.51 |
|
A+B |
12.3 |
0.04 |
0.78 |
3.9 |
0.05 |
0.66 |
0.6 |
0.01 |
0.57 |
|
A+B+C |
13.9 |
0.04 |
0.79 |
4.2 |
0.05 |
0.67 |
0.7 |
0.01 |
0.58 |
| A+B+C+D | 14.0 | 0.04 | 0.79 | 6.7 | 0.05 | 0.70 | 0.8 | 0.01 | 0.58 |
Modeling the c-Jun binding to AP-1 motif (TGAC/GTCA), GR binding to GR motif (G-ACA---TGT-C), Hoxa2 binding to Pbx motif (TGATTGAT) were performed by the logistic regression using generalized linear models (GLM). PVE Percent of variance explained (%) = 1-(deviance/null.deviance). CV err. estimated 11-fold cross-validation prediction error, AUC Area under the ROC curve. *The Co-occurrence of 2nd motifs: the co-occurrence motif of AP-1 motif is GR motif (G-ACA---TGT-C), the co-occurrence of GR motif is AP-1 motif (TGAC/GTCA), the co-occurrence of Pbx motif is GR-like 8-mer (G-TGA---ATG-C). The PVEs of INOS and Co-occurrence of 2nd motif for AP-1, GR and Pbx motif are shown in bold to highlight the difference between the motifs.