BMC Bioinformatics

Table 1 Performance of DNN models for predicting eight datasets of kinase family inhibitors

From: Exploring kinase family inhibitors and their moiety preferences using deep SHapley additive exPlanations

Group	Family	Test set			r^b
Group	Family	ACC (avg.)^a	AUC (avg.)^a	MCC (avg.)^a	r^b
TK	EGFR	0.93	0.96	0.85	0.72
	Jak	0.92	0.96	0.83	0.45
CMGC	GSK	0.79	0.77	0.45	0.62
	CLK	0.75	0.74	0.37	0.12
CAMK	PIM	0.91	0.95	0.82	0.75
	PKD	0.85	0.68	0.28	0.18
AGC	Akt	0.93	0.98	0.85	0.86
	PKG	0.88	0.82	0.47	0.52

^aThe average performance of predictions by 100 models on validation sets
^bThe Pearson’s correlation coefficient between the SHAP scores and the odds ratios of the top 15 moiety features for each kinase family (refer to Methods: Features of compound datasets)

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com