Lipinski's Rule of Five and Druglikeness Done Right

Where the four numbers came from

In 1997 Christopher Lipinski and colleagues looked back over thousands of orally active compounds and noticed they clustered. From that pattern came the rule of five: poor oral absorption becomes likely when a molecule breaks more than one of four limits. The "five" in the name is because every threshold is a multiple of five.

Lipinski's Rule of Five (a flag is raised if you break >1):

  Molecular weight        <= 500
  logP                    <= 5
  H-bond donors           <= 5   (count of N-H and O-H)
  H-bond acceptors        <= 10  (count of N and O)

Note what it covers: passive oral ABSORPTION only.
Note what it ignores: potency, selectivity, metabolism,
                      toxicity, and active transport.

Each number traces back to properties you now understand. The molecular weight cap reflects that bigger molecules diffuse and dissolve more slowly. The logP cap is the upper edge of the lipophilicity sweet spot. The donor and acceptor limits are stand-ins for polar surface area and the cost of shedding water to cross a membrane. The rule is really four familiar properties wearing round-number clothes.

What it does — and does not — tell you

The rule of five is a risk flag for passive oral absorption, nothing more. It says nothing about whether a molecule is potent, selective, metabolically stable, or safe. A compound can pass all four limits and still be a useless drug; a compound can break a limit and be a blockbuster. Lipinski himself framed it as a guideline, not a law — and explicitly noted exceptions.

The exceptions are instructive. Whole drug classes live happily outside the box: natural products, macrocycles, and many antibiotics are large and polar yet orally active, often because they ride on active transporters or adopt folded shapes that hide their polarity. The rule was built on passive diffusion, so it simply does not apply where another mechanism carries the molecule across. Use it to raise an eyebrow, not to slam a door.

From a single gate to multiparameter design

Modern teams have moved past pass/fail. The reality is that no single property decides a drug; you must optimize many at once — potency, druglikeness, solubility, permeability, metabolic stability, and safety — and they conflict. This is multiparameter optimization, the daily craft of property-based design. The rule of five becomes one input among many, not the verdict.

A healthier lens than "did it pass?" is efficiency. Ligand efficiency asks how much potency you earn per heavy atom, discouraging bloat; lipophilic ligand efficiency asks how much you earn per unit of greasiness. Tracking these from the hit stage keeps a series lean, so that when it grows during optimization it still lands inside druglike space rather than drifting out of it.

1. Compute the rule-of-five properties early — they are free and flag obvious risk.
2. Treat any break as a question, not a rejection: is there a transporter, a folded shape, or a non-oral route that makes it fine?
3. Watch the trend across a series — if weight and logP creep up every cycle, you are heading out of druglike space.
4. Optimize potency, properties, and safety together — let multiparameter scoring, not one rule, guide which compound to make next.