Table 1.

Performance comparison of four competing methods on 1000 gene datasets and 500 gene datasets.

EPR
Methods	1000 Gene Datasets							500 Gene Datasets
	mHSC-L	mHSC-GM	mHSC-E	hESC	mESC	hHep	mDC	mHSC-L	mHSC-GM	mHSC-E
DeepSEM	1.09	1.14	1.24	1.43	1.06	1.14	2.55	1.07	1.06	1.30
DGRN	-	-	-	-	-	-	-	-	-	-
GENIE3	<1	1.03	1.01	1.00	1.06	1.12	1.01	1.00	1.06	1.03
PIDC	<1	<1	<1	<1	1.01	1.03	<1	<1	<1	<1
MetaSEM	1.36	1.41	1.24	2.29	1.24	1.53	3.36	1.13	1.20	1.69
AUPR
Methods	1000 Gene Datasets							500 Gene Datasets
	mHSC-L	mHSC-GM	mHSC-E	hESC	mESC	hHep	mDC	mHSC-L	mHSC-GM	mHSC-E
DeepSEM	0.63	0.56	0.42	0.30	0.38	0.48	0.43	0.66	0.63	0.58
DGRN	0.15	0.16	0.25	-	-	-	-	0.15	0.27	0.25
GENIE3	0.09	0.12	0.09	-	-	-	-	0.14	0.15	0.14
PIDC	0.07	0.12	0.10	-	-	-	-	0.16	0.12	0.19
MetaSEM	0.70	0.73	0.66	0.48	0.43	0.70	0.33	0.75	0.77	0.84
AUROC
Methods	1000 Gene Datasets							500 Gene Datasets
	mHSC-L	mHSC-GM	mHSC-E	hESC	mESC	hHep	mDC	mHSC-L	mHSC-GM	mHSC-E
DeepSEM	0.51	0.57	0.63	0.52	0.51	0.54	0.77	0.52	0.52	0.67
DGRN	0.63	0.67	0.75	-	-	-	-	0.63	0.71	0.77
GENIE3	0.63	0.63	0.59	-	-	-	-	0.52	0.52	0.55
PIDC	0.57	0.61	0.60	-	-	-	-	0.47	0.47	0.59
MetaSEM	0.75	0.77	0.76	0.81	0.61	0.77	0.71	0.67	0.72	0.87

Part of the results are from the following two papers [10,15]; “-” indicates that the experimental result is missing, and “<1” indicates that the result here is random prediction. The bold number refers to the best performer.