PaPI: pseudo amino acid composition to score human protein-coding variants

BMC Bioinformatics

Table 4 Performances of RF, PolyPhen2, SIFT and PaPI on the three test sets

Test Set	Tool	AUC	Accuracy [IC95%]	Sens	Spec	PPV	NPV	F-m	MCC
# 1	PaPI	.9207	.8621 [.8553-.8685]	.8580	.8663	.8688	.8553	.8633	.7242
	RF	.8941	.8262 [.8189-.8334]	.8286	.8238	.8291	.8233	.8289	.6524
	PolyPhen2	.9137	.8425 [.8354-.8493]	.8533	.8314	.8392	.846	.8462	.6849
	SIFT	.8682	.8045 [.7968-.812]	.7724	.8376	.8307	.781	.8005	.6108
# 2	PaPI	.9196	.8618 [.8550-.8683]	.8572	.8665	.869	.8545	.8631	.7236
	RF	.8960	.8275 [.8202-.8346]	.8319	.823	.8292	.8257	.8305	.6549
	PolyPhen2	.9121	.8401 [.8330-.847]	.8486	.8314	.8387	.8417	.8436	.6801
	SIFT	.8677	.7994 [.7917-.807]	.7625	.8376	.829	.7735	.7944	.6013
# 3	PaPI	.9239	.8648 [.8568-.8724]	.8570	.8729	.8745	.8553	.8657	.7298
	RF	.8999	.8289 [.8202-.8373]	.8358	.8218	.8289	.8289	.8323	.6577
	PolyPhen2	.9185	.8416 [.8331-.8497]	.8501	.8328	.8401	.8432	.8451	.6831
	SIFT	.8688	.7999 [.7906-.8088]	.7558	.8454	.8348	.7701	.7933	.603

Performances of the Random Forest (RF), PolyPhen2, SIFT and PaPI (RF + PolyPhen2 + SIFT) on the three test. Area under the curve (AUC), accuracy with 95% confidence interval, sensitivity (Sens), specificity (Spec), Positive Predictive Value (PPV), Negative Predictive Value (NPV), F-measure (F-m) and Matthews correlation coefficient (MCC) are reported for each method. Test sets were filtered in order to retain only those variants that both PolyPhen2 and SIFT were able to predict.

ISSN: 1471-2105