Fast Bayesian inference for gene regulatory networks using ScanBMA

BMC Systems Biology

Table 1 Performance of different methods on the yeast data

Method	Precision	AUROC	AUPRC	TP	FP
LASSO	0.046	0.506	0.0416	996	20,469
ARACNE	0.205	0.502	0.0399	69	268
CLR	0.039	0.510	0.0435	8,879	220,942
MRNET	0.039	0.513	0.0442	8,737	214,757
ScanBMA ^[20]	0.391	0.601	0.0747	227	353
ScanBMA ^[3556]	0.274	0.629	0.0740	127	336
iBMA ^[100]	0.180	0.517	0.0788	593	2,702

AUROC is the area under the ROC curve, AUPRC is the area under the precision-recall curve, and TP and FP are the numbers of true positive and false positive edges inferred, respectively. Thus TP +FP is the number of edges in the inferred network and Precision = TP/(TP +FP). ScanBMA was applied to the transformed data using the informative edge prior and Zellner’s g-prior for the model parameters. The superscript indicates the value of nvar. Expected precision and AUPRC from random guessing is 0.0380.

ISSN: 1752-0509