Skip to main content

Table 3 Performances of classification and regression bioactivity models for the A1AR and A2AAR

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

ProteinModel training typeDataset in trainingValidation set (only A1AR or A2AAR compounds, respective of the protein)MCCSensitivitySpecificityPPVNPVROC
A1ARClassificationA1AR compounds in the A1AR/A2AAR dataset + semi-selective compoundsA1AR/A2AAR dataset + semi-selective compounds− 0.09 ± 0.060.44 ± 0.090.48 ± 0.110.46 ± 0.090.45 ± 0.060.41 ± 0.05
ClassificationA1AR bioactivity datasetA1AR/A2AAR dataset + semi-selective compounds− 0.16 ± 0.050.39 ± 0.080.45 ± 0.100.42 ± 0.090.41 ± 0.060.39 ± 0.04
RegressionA1AR compounds in the A1AR/A2AAR dataset + semi-selective compoundsA1AR/A2AAR dataset + semi-selective compounds0.09 ± 0.040.53 ± 0.090.56 ± 0.080.54 ± 0.090.54 ± 0.080.61 ± 0.03
RegressionA1AR bioactivity datasetA1AR/A2AAR dataset + semi-selective compounds0.04 ± 0.060.46 ± 0.080.58 ± 0.080.52 ± 0.100.52 ± 0.080.59 ± 0.05
RegressionA1AR bioactivity datasetA1AR bioactivity dataset0.06 ± 0.070.49 ± 0.070.58 ± 0.080.53 ± 0.070.54 ± 0.060.60 ± 0.05
A2AARClassificationA2AAR compounds in the A1AR/A2AAR dataset + semi-selective compoundsA1AR/A2AAR dataset + semi-selective compounds0.11 ± 0.090.59 ± 0.100.50 ± 0.130.73 ± 0.050.39 ± 0.050.59 ± 0.06
ClassificationA2AAR bioactivity datasetA1AR/A2AAR dataset + semi-selective compounds0.16 ± 0.110.57 ± 0.120.56 ± 0.130.75 ± 0.060.45 ± 0.090.61 ± 0.07
RegressionA2AAR compounds in the A1AR/A2AAR dataset + semi-selective compoundsA1AR/A2AAR dataset + semi-selective compounds0.12 ± 0.100.69 ± 0.100.40 ± 0.080.70 ± 0.040.47 ± 0.100.64 ± 0.06
RegressionA2AAR bioactivity datasetA1AR/A2AAR dataset + semi-selective compounds0.21 ± 0.070.64 ± 0.100.56 ± 0.100.76 ± 0.040.46 ± 0.040.69 ± 0.05
RegressionA2AAR bioactivity datasetA2AAR bioactivity dataset0.19 ± 0.070.63 ± 0.110.54 ± 0.090.70 ± 0.040.50 ± 0.050.69 ± 0.05
  1. Query compounds were categorized based on post-classification of the bioactivity predictions: predicted pActivity < 6.5 = inactive and predicted pActivity ≥ 6.5 = active
  2. MCC Matthews Correlation Coefficient, PPV positive predictive value, NPV negative predictive value, ROC receiver operating characteristic