Skip to main content

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

From: Quantitative prediction of selectivity between the A1 and A2A adenosine receptors

Model ClassMCCSensitivitySpecificityPPVNPVROC
Selectivity-windowA1AR0.130.390.830.120.960.75
A2AAR0.400.241.000.700.970.66
Dual0.020.810.210.640.390.37
Selectivity-window and bioactivity filteredA1AR0.181.000.160.211.000.66
A2AAR0.881.000.970.801.000.98
Dual0.001.000.280.72
  1. The query compounds were categorized based on post hoc optimized classification of the selectivity predictions: A1AR-selective when selectivity-window ≥ 0.5, A2AAR-selective when selectivity-window ≤ − 2, and dual binder when selectivity-window ≥ − 1 and < 0.5
  2. MCC Matthews Correlation Coefficient, PPV positive predictive value, NPV negative predictive value, ROC receiver operating characteristic