Overview of the multi-objective optimisation benchmarks used in this paper, the first five benchmarks are adapted from the Guacamol suite while the latter two benchmarks were constructed to emulate the demands of poly-pharmacology projects. The upper row of each task represents the values calculated for each objective. The lower rows show the modifiers applied to each of these values. The fingerprints used to calculate the similarities are denoted as arguments of the Tanimoto function, the parameters used for the modifiers are displayed as arguments of the corresponding functions. For the poly-pharmacology benchmarks, the genes targeted for activity are indicated. The CNS function calculates the central nervous system desirability score (high blood–brain-barrier permeability and low toxicity potential) as proposed by Pfizer51.
| Task\objective | I | II | III | IV | V |
|---|---|---|---|---|---|
| Cobimetinib | |||||
| Tanimoto(FCFP4) | Tanimoto(ECFP6) | Rotatable bonds | Aromatic rings | CNS(0.5) | |
| Clipped(0.7) | MinGaussian(0.75, 0.1) | MinGaussian(3, 1) | MaxGaussian(3, 1) | — | |
| Fexofenadine | |||||
| Tanimoto(AP) | TPSA | log(P) | — | — | |
| Clipped(0.8) | MaxGaussian(90, 10) | MinGaussian(4, 1) | — | — | |
| Osimertinib | |||||
| Tanimoto(FCFP4) | Tanimoto(ECFP6) | TPSA | log(P) | — | |
| Clipped(0.8) | MinGaussian(0.85, 0.1) | MaxGaussian(95, 20) | MinGaussian(1, 1) | — | |
| Pioglitazone | |||||
| Tanimoto(ECFP4) | Molecular weight | Rotatable bonds | — | — | |
| Gaussian(0, 0.1) | Gaussian(356, 10) | Gaussian(2, 0.5) | — | — | |
| Ranolazine | |||||
| Tanimoto(AP) | log(P) | TPSA | Fluorine count | — | |
| Clipped(0.7) | MaxGaussian(7, 1) | MaxGaussian(95, 20) | Gaussian(1, 1) | — | |
| DAP kinases | |||||
| hERG | SCN2A | DAPk1 | DRP1 | ZIPk | |
| Gaussian(0, 0.1) | Gaussian(0, 0.1) | Clipped(0.8) | Clipped(0.8) | Clipped(0.8) | |
| Antipsychotics | |||||
| hERG | 5-HT2A | 5-HT2B | DRD2 | CNS(0.5) | |
| Gaussian(0, 1.0) | Clipped(0.8) | Clipped(0.8) | Clipped(0.8) | — | |