Documentation
Druglikeness scoring
Not every molecule that binds a target will make a good drug. It also needs favorable pharmacokinetic properties — absorption, distribution, metabolism, and excretion (ADME). OpenDDE computes druglikeness scores via RDKit.
Lipinski’s Rule of Five
The most widely used druglikeness filter, proposed by Christopher Lipinski in 1997. A compound is likely orally bioavailable if it does not violate more than one of these rules:
| Property | Threshold | Why it matters |
|---|---|---|
| Molecular weight | ≤ 500 Da | Larger molecules have trouble crossing cell membranes |
| LogP | ≤ 5 | Too lipophilic = poor solubility and high toxicity risk |
| H-bond donors | ≤ 5 | Too many donors reduce membrane permeability |
| H-bond acceptors | ≤ 10 | Too many acceptors reduce membrane permeability |
Veber’s rules
Additional filters for oral bioavailability:
- TPSA (topological polar surface area) ≤ 140 Ų
- Rotatable bonds ≤ 10
Computed properties
For every ligand, OpenDDE computes via RDKit:
- Molecular weight
- LogP (Wildman-Crippen method)
- Hydrogen bond donors and acceptors
- TPSA
- Rotatable bonds count
- Number of Lipinski violations