Quantitative prediction of selectivity between the A1 and A2A adenosine receptors

Table 5 Performance of the selectivity-window model on an external validation set

Model	Class	MCC	Sensitivity	Specificity	PPV	NPV	ROC
Selectivity-window	A₁AR	0.13	0.39	0.83	0.12	0.96	0.75
	A_2AAR	0.40	0.24	1.00	0.70	0.97	0.66
	Dual	0.02	0.81	0.21	0.64	0.39	0.37
Selectivity-window and bioactivity filtered	A₁AR	0.18	1.00	0.16	0.21	1.00	0.66
	A_2AAR	0.88	1.00	0.97	0.80	1.00	0.98
	Dual	–	0.00	1.00	–	0.28	0.72

The query compounds were categorized based on post hoc optimized classification of the selectivity predictions: A₁AR-selective when selectivity-window ≥ 0.5, A_2AAR-selective when selectivity-window ≤ − 2, and dual binder when selectivity-window ≥ − 1 and < 0.5
MCC Matthews Correlation Coefficient, PPV positive predictive value, NPV negative predictive value, ROC receiver operating characteristic

ISSN: 1758-2946