Identifying novel host-based diagnostic biomarker panels for COVID-19: a whole-blood/nasopharyngeal transcriptome meta-analysis

Table 4 The comparison of criteria related to the best panels of the LASSO feature-based prediction model and RF-based generic prediction model

Tissue	Number of features	The best LASSO-based panels based on train and test sets			The best common-based panels based on train and test sets			Difference of accuracies
Tissue	Number of features	Sensitivity	Specificity	Accuracy	Sensitivity	Specificity	Accuracy	Difference of accuracies
WB	3	0.906	0.923	0.911	0.938	0.769	0.889	0.022
	4	0.906	0.923	0.911	0.969	0.692	0.889	0.022
	5	0.906	0.923	0.911	0.969	0.769	0.911	0.000
	6	1	0.923	0.978	0.938	0.846	0.911	0.067
	7	1	0.923	0.978	0.938	0.846	0.911	0.067
	8	1	0.923	0.978	0.938	0.846	0.911	0.067
	9	1	0.923	0.978	0.906	0.846	0.889	0.089
NP	3	0.816	0.786	0.808	0.743	0.802	0.758	0.050
	4	0.856	0.825	0.848	0.810	0.810	0.810	0.038
	5	0.864	0.873	0.866	0.826	0.841	0.830	0.036
	6	0.853	0.857	0.854	0.829	0.865	0.838	0.016
	7	0.869	0.873	0.870	0.832	0.881	0.844	0.026
	8	0.872	0.913	0.882	0.850	0.841	0.848	0.034
	9	0.872	0.905	0.880	0.850	0.857	0.852	0.028

ISSN: 1528-3658