Genomic Promoter Analysis Predicts Functional Transcription Factor Binding

Table 2

Training set self-consistency performance. Data from ChIP-chip studies (3) were integrated with CONFAC TFBS data and genes were separated randomly into training and test sets. The BAM MEI classifier was trained on the training set of 5399 genes and predictions were made on this same set of genes.

Predicted
Observed	HNF1 both	HNF1 hep	HNF1 panc	HNF4 both	HNF4 hep	HNF4 panc	Unbound	Total (obs)	Sensitivity
HNF1 both	10	0	0	0	0	0	4	14	71%
HNF1 hep	0	4	0	0	0	0	74	78	5%
HNF1 panc	0	0	10	0	0	0	14	24	41%
HNF4 both	0	0	0	130	0	0	142	272	48%
HNF4 hep	0	0	0	0	88	0	295	383	23%
HNF4 panc	0	0	0	0	0	181	159	340	53%
Unbound	0	0	0	6	5	2	4249	4262	99%
total (pred)	10	4	10	136	93	183	4937	5373
Specificity	100%	NA	100%	89%	85%	96%	86%