Integrative Deep Learning for Identifying Differentially Expressed (DE) Biomarkers

Table 1

Simulation results of metalogistic regression, meta-SVM, and the integrative deep learning (no inclusion of random study).

Methods	σ	Sensitivity (s.e)		Specificity (s.e)		Youden J index
Meta-SVM	0.1	1	(0)	0.9859	(0.001)	0.9859
	0.3	1	(0)	0.1597	(0.007)	0.1597
	0.5	0.9930	(0.0006)	0.1498	(0.006)	0.1491
	0.7	0.9777	(0.0003)	0.5399	(0.007)	0.5100
	0.9	0.9999	(0)	0.0135	(0.001)	0.0120
Metalogistic regression	0.1	0.068	(0.0017)	1	(0)	0.0680
	0.3	0.2000	(0.006)	1	(0)	0.2000
	0.5	0.2937	(0.0089)	0.9814	(0.0029)	0.2715
	0.7	0.3027	(0.01)	0.9484	(0.005)	0.2512
	0.9	0.3006	(0.012)	0.9085	(0.005)	0.2092
Integrative deep learning	0.1	0.7502	(0.02)	0.845312	(0.02)	0.5973
	0.3	0.6625	(0.026)	0.7817	(0.02)	0.4427
	0.5	0.7208	(0.024)	0.6911	(0.024)	0.4119
	0.7	0.7042	(0.026)	0.6625	(0.022)	0.3666
	0.9	0.7000	(0.03)	0.6427	(0.029)	0.3427