Lipinski Rule of Five Explained: A Worked Pass/Fail Walkthrough
Lipinski Rule of Five explained: the four thresholds, a worked aspirin and atorvastatin pass/fail, and when one violation is fine.
You draw an analog, the property panel updates, and a red badge appears: Lipinski: 1 violation. The instinct is to treat that like a failing grade and redraw. It usually isn’t one. The Lipinski Rule of Five was never a pass/fail gate — it is a probabilistic flag for passive oral absorption, and plenty of marketed oral drugs trip a criterion and ship anyway. This post walks the four rules with their actual thresholds, evaluates two real drugs against them by hand, and shows where the rule quietly misleads you.
The Lipinski Rule of Five explained: the four criteria
Christopher Lipinski and colleagues at Pfizer published the rule in 1997, derived from the properties of compounds that had reached Phase II oral trials. The name is a mnemonic: every cutoff is a multiple of five.
- Molecular weight — MW < 500 Da
- Lipophilicity — logP < 5 (octanol–water partition coefficient)
- Hydrogen bond donors — HBD < 5 (count of N–H and O–H)
- Hydrogen bond acceptors — HBA < 10 (in Lipinski’s original count, the sum of nitrogen and oxygen atoms)
The decision rule is the part people forget: a compound is flagged only when it violates two or more criteria. One violation is within the rule, not against it. A single red marker on a property panel is a prompt to look closer, not a verdict.
A worked pass/fail evaluation: aspirin and atorvastatin
Take two molecules a chemist already trusts and run them against the four rules. Both are orally dosed; one clears every criterion, one trips a cutoff and is still a blockbuster oral drug.
- MW = 180.16 Da — under 500. Pass.
- logP ≈ 1.2 (Wildman–Crippen) — under 5. Pass.
- HBD = 1 (the carboxylic acid O–H) — under 5. Pass.
- HBA = 4 (four oxygen atoms, Lipinski count) — under 10. Pass.
- MW = 558.6 Da — over 500. Violation.
- logP ≈ 4.1 — under 5. Pass.
- HBD = 4 (two hydroxyls, one amide N–H, one carboxylic acid O–H) — under 5. Pass.
- HBA = 7 (two nitrogens + five oxygens, Lipinski count) — under 10. Pass.
This is the practical reading of the badge. A pass-with-one-violation molecule like atorvastatin should not be deprioritized on Ro5 alone. The rule earns its keep by catching compounds that miss several cutoffs at once — those are the ones whose oral absorption is genuinely at risk.
When the Rule of Five misleads
Ro5 predicts passive oral absorption for a particular slice of chemical space. Outside that slice it goes quiet or wrong, and a working chemist needs to know the exceptions before trusting the badge.
- Actively transported drugs. Molecules taken up by transporters (many antibiotics, some kinase inhibitors) don’t rely on passive diffusion, so a Ro5 violation doesn’t predict poor uptake.
- Natural products. Macrolides and many natural-product scaffolds violate multiple criteria and are still orally active — Lipinski himself excluded natural products and substrates of biological transporters from the rule’s scope.
- Non-oral routes. Injectables, inhaled, and topical drugs aren’t bound by an oral-absorption heuristic at all. Applying Ro5 to an IV candidate is a category error.
- It says nothing about potency or toxicity. Ro5 is silent on whether the molecule binds the target or is safe. A compound can pass all four rules and be inactive.
For oral compounds that sit at the edge of Ro5, chemists usually reach for the complementary TPSA permeability thresholds and Veber criteria — the Veber rules add rotatable bonds (≤ 10) and topological polar surface area (≤ 140 Å2) as absorption predictors that Ro5 omits. Reading those alongside the four classic rules gives a fuller picture than the badge alone.
Reading the property panel in practice
When you draw a structure in ChemStitch, the property panel computes MW, logP (Wildman–Crippen), HBD, HBA, TPSA, and rotatable bonds from RDKit and shows a Lipinski pass/fail badge with the violation count. The badge counts violations exactly as above: two or more flips it to fail. The values are marked Computed (deterministic, green), distinct from any AI Suggested value — which matters when you’re deciding whether to trust a number near a cutoff. If you need to confirm a molecular weight by hand before reading the badge, the molecular weight calculator gives you the MW directly from a structure or formula. For the salt-form correction that trips up MW comparisons, see free-base versus salt-form MW correction.
The badge is a starting filter, not a gate. Read it the way Lipinski intended: a two-violation flag is worth investigating, a one-violation molecule is still in the running, and an injectable or actively-transported compound shouldn’t be judged by it at all.
References: Lipinski et al., “Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings,” Adv. Drug Deliv. Rev. (DOI: 10.1016/S0169-409X(00)00129-0, PubMed); Veber et al., “Molecular properties that influence the oral bioavailability of drug candidates,” J. Med. Chem. 2002 (DOI: 10.1021/jm020017n); RDKit Crippen logP and descriptor documentation (rdkit.org).