Optimally choosing PWM motif databases and sequence scanning approaches based on ChIP-seq data

BMC Bioinformatics

Table 4 Average balanced accuracy for optimal threshold parameters

Threshold selection	Average BA for parameter:
method
	Optimal constant	Fitted with linear model	Peaks’ flanks	3rd exons	Shuffled peaks	3rd order MCs
Log-likelihood	0.6704	0.6724	8.9	8	9	9.9
FPR	0.6749	0.6757	0.00025	0.00041	0.00021	0.00013
FNR	0.6285	0.6641	0.26	0.18	0.27	0.32
Balanced	0.6664	0.6743	794	457	977	2089

Only motifs with AUC>0.6 are considered. Balanced accuracies are calculated with respect to negative sequences composed of flanks of ChIP-seq peaks. 2nd column contains average balanced accuracies for thresholds in 4th column. 3rd column contains average balanced accuracies for thresholds linearly depending on the motif features: information content, log-likelihood value range, log-likelihood standard deviation.

ISSN: 1471-2105