NCBoost classifies pathogenic non-coding variants in Mendelian diseases through supervised learning on purifying selection signals in humans

Table 2 AUROC and AUPRC values obtained by NCBoost upon different configurations of the training and independent testing sets. The figures obtained by the six state-of-the-art methods evaluated on the same testing sets are shown together with NCBoost

Positive set used for NCBoost training		Positive set used for testing		AUROC							AUPRC
Source	# SNVs	Source	# SNVs	CADD	DeepSEA	Eigen	Eigen PC	FunSeq2	ReMM	NCBoost	CADD	DeepSEA	Eigen	Eigen PC	FunSeq2	ReMM	NCBoost
HGMD-DM	186	!HGMD-DM	97	0.67	0.76	0.78	0.70	0.67	0.74	0.82	0.16	0.24	0.23	0.13	0.15	0.26	0.36
CV	107	!CV	176	0.72	0.74	0.73	0.63	0.68	0.77	0.78	0.27	0.31	0.24	0.11	0.15	0.37	0.38
Smedley	78	!Smedley	205	0.69	0.72	0.73	0.65	0.66	0.75	0.78	0.24	0.25	0.21	0.11	0.14	0.31	0.36
≥ 2 sources	73	1 source	210	0.68	0.72	0.73	0.64	0.66	0.75	0.78	0.23	0.24	0.21	0.11	0.14	0.31	0.31
1 source	210	≥ 2 sources	73	0.8	0.81	0.82	0.69	0.77	0.82	0.85	0.31	0.43	0.29	0.16	0.24	0.48	0.52

ISSN: 1474-760X