A scalable algorithm for structure identification of complex gene regulatory network from temporal expression data

Table 2 Evaluation of DMI at a noise level from 10 to 30% for a network size 10, 100, 1,000 or 20,000, based on the arithmetic average over 10 simulation runs

Noise level	Size	SN	SP	ACC	MCC	AUC
10% noise	10	0.9333 ± 0.0573	0.9000 ± 0.0860	0.9200 ± 0.0688	0.8333 ± 0.1434	0.9166 ± 0.0717
	100	0.7687 ± 0.0379	0.8253 ± 0.0407	0.7984 ± 0.0283	0.5965 ± 0.0571	0.7970 ± 0.0282
	1000	0.8273 ± 0.0244	0.8305 ± 0.0131	0.8289 ± 0.0182	0.6579 ± 0.0363	0.8289 ± 0.0182
	20k	0.7748 ± 0.0041	0.7753 ± 0.0041	0.7751 ± 0.0041	0.5502 ± 0.0083	0.7751 ± 0.0041
20% noise	10	0.8000 ± 0.1365	0.7000 ± 0.2048	0.7600 ± 0.1639	0.5000 ± 0.3414	0.7500 ± 0.1707
	100	0.7312 ± 0.0440	0.7569 ± 0.0398	0.7447 ± 0.0418	0.4881 ± 0.0838	0.7440 ± 0.0419
	1000	0.8056 ± 0.0133	0.8006 ± 0.0137	0.8031 ± 0.0135	0.6063 ± 0.0271	0.8031 ± 0.0135
	20k	0.7452 ± 0.0045	0.7457 ± 0.0045	0.7454 ± 0.0045	0.4909 ± 0.0091	0.7454 ± 0.0045
30% noise	10	0.7333 ± 0.1194	0.6000 ± 0.1791	0.6800 ± 0.1433	0.3333 ± 0.2986	0.6667 ± 0.1493
	100	0.7005 ± 0.0320	0.7291 ± 0.0289	0.7155 ± 0.0304	0.4297 ± 0.0610	0.7148 ± 0.0305
	1000	0.7829 ± 0.0080	0.7774 ± 0.0082	0.7802 ± 0.0081	0.5603 ± 0.0162	0.7801 ± 0.0081
	20k	0.7339 ± 0.0048	0.7344 ± 0.0048	0.7342 ± 0.0048	0.4684 ± 0.0096	0.7342 ± 0.0048

ISSN: 1471-2105