Summary comparison of accuracy of drug interactions prediction for CYP3A using testosterone, midazolam, and felodipine activity assays
[I] = Csys | [I] = fu × Chep,inlet(est) | |
|---|---|---|
| Testosterone 6β-hydroxylase | ||
| Geometric mean-fold error (All DDI) | 1.47 | 1.45 |
| Geometric mean-fold error (DDI ≥2× only) | 2.00 | 1.89 |
| RMSE (All DDI) | 6.87 | 2.18 |
| RMSE (DDI ≥2× only) | 11.12 | 3.50 |
| Identification of true positives (DDI ≥2×) | 10 of 16 | 9 of 16 |
| Identification of true negatives (DDI <2×) | 26 of 26 | 25 of 26 |
| Overall Success Rate (%)a | 86 | 81 |
| Midazolam 1′-hydroxylase | ||
| Geometric mean-fold error (All DDI) | 1.44 | 1.48 |
| Geometric mean-fold error (DDI ≥2× only) | 1.91 | 1.99 |
| RMSE (All DDI) | 4.90 | 2.95 |
| RMSE (DDI ≥2× only) | 7.93 | 4.74 |
| Identification of true positives (DDI ≥2×) | 12 of 16 | 11 of 16 |
| Identification of true negatives (DDI <2×) | 26 of 26 | 25 of 26 |
| Overall Success Rate (%)a | 90 | 86 |
| Felodipine dehydrogenase | ||
| Geometric mean-fold error (All DDI) | 1.52 | 1.53 |
| Geometric mean-fold error (DDI ≥2× only) | 2.19 | 2.16 |
| RMSE (All DDI) | 6.27 | 2.76 |
| RMSE (DDI ≥2× only) | 10.15 | 4.44 |
| Identification of true positives (DDI ≥2×) | 10 of 16 | 8 of 16 |
| Identification of true negatives (DDI <2×) | 26 of 26 | 25 of 26 |
| Overall success rate (%)a | 86 | 79 |
| Most potently inhibited activity | ||
| Geometric mean-fold error (All DDI) | 1.44 | 1.44 |
| Geometric mean-fold error (DDI ≥2× only) | 1.91 | 1.87 |
| RMSE (All DDI) | 7.25 | 2.57 |
| RMSE (DDI ≥2× only) | 11.74 | 4.13 |
| Identification of true positives (DDI ≥2×) | 12 of 16 | 11 of 16 |
| Identification of true negatives (DDI <2×) | 26 of 26 | 25 of 26 |
| Overall success rate (%)a | 90 | 86 |
↵ a The term “Overall Success Rate” refers to the percentage of times that predictions of drug interactions from in vitro data are properly categorized as greater than or less than 2-fold