Table 2.

Performances of three proposed models in datasets GSE29676 and GSE39087.

	Average LOOCV accuracy	Validation accuracy	Sensitivity	Specificity	NPV	PPV	FDR	FOR
Cross-validation in GSE29676
ECH1 + NHLRC2	98.8%	95.4%	94.0%	97.1%	93.0%	97.7%	2.3%	7.0%
ECH1 + HOXB7	97.7%	95.6%	95.0%	96.3%	94.1%	97.1%	2.9%	5.9%
ERBB2 + FN1 + SLC6A13	96.5%	89.6%	87.9%	91.8%	86.4%	93.4%	6.6%	13.6%
The Nagele model	64.0%	56.8%	58.9%	54.3%	51.8%	62.2%	37.8%	48.2%
Cross-validation in GSE39087
ECH1 + NHLRC2	88.9%	87.4%	79.7%	90.8%	91.2%	80.7%	19.4%	8.8%
ECH1 + HOXB7	97.8%	96.9%	96.8%	96.9%	98.6%	93.8%	6.2%	1.4%
ERBB2 + FN1 + SLC6A13	74.4%	69.8%	80.9%	65.0%	89.4%	51.4%	48.6%	10.6%
The Nagele model	70.0%	69.4%	80.2%	64.7%	89.0%	50.6%	49.4%	11.0%