Joint estimation of causal effects from observational and intervention gene expression data

BMC Systems Biology

Table 4 Comparison of methods on partial DREAM4 data

Criterion	Dataset	MCMC-	Pinna	IDA	IDA
		Mallows		(opt)	(pes)
	1	0.708	0.555	0.448	0.413
	2	0.525	0.637	0.610	0.641
AUROC	3	0.711	0.498	0.638	0.638
	4	0.748	0.682	0.554	0.483
	5	0.676	0.565	0.599	0.534
	1	0.344	0.346	0.142	0.133
	2	0.240	0.306	0.243	0.284
AUPRC	3	0.271	0.214	0.242	0.242
	4	0.322	0.494	0.162	0.158
	5	0.194	0.168	0.146	0.156
DREAM score	overall	1.844	1.450	0.735	0.723

Area under the ROC curve (AUROC), area under the precision-recall curve (AUPRC), and DREAM score for each of the five DREAM4 partial datasets, where only five of the single-gene knock-outs are included, for the MCMC-Mallows, Pinna et al., and IDA (optimistic and pessimistic) methods. Results for the Petri Nets method [17] are not provided as no software is publicly available to implement this approach. Evaluation scripts for the overall DREAM score were obtained from the DREAM4 evaluation page, located at http://wiki.c2b2.columbia.edu/dream/results/DREAM4.

ISSN: 1752-0509