Rule of Five, Veber, Ghose, Egan

Filters are how a hit list of a hundred thousand becomes a shortlist of fifty. Used well they are a defensible triage; used badly they delete good compounds and teach you nothing. The difference is entirely in knowing what question each filter was originally asking.

Lipinski and colleagues looked at compounds that had reached phase II — that is, molecules that had already survived enough to be dosed in patients — and asked what they had in common. Poor absorption or permeation is more likely when:

• molecular weight is over 500;
• cLogP is over 5;
• there are more than 5 hydrogen bond donors;
• there are more than 10 hydrogen bond acceptors, counted as nitrogen plus oxygen.

Three things are routinely forgotten about it, and each matters.

• It was a statement about oral absorption, not about whether a molecule is any good. A compound that fails it may be an excellent injectable, an excellent tool compound, or an excellent starting point.
• It allows one violation. Two is the flag. It is often applied as though any violation were disqualifying.
• The acceptor count is the crude one — every N and O, as module 12 explained. Applying the threshold to a chemically-defined acceptor count is applying it to a different quantity.

Several filters followed, each adding something the Rule of Five missed.

• Veber — rotatable bonds ≤ 10 and TPSA ≤ 140 Ų. Adds flexibility and polar surface, neither of which the Rule of Five measures at all. Derived from rat oral bioavailability data.
• Ghose — molecular weight 160 to 480, cLogP −0.4 to 5.6, molar refractivity 40 to 130, 20 to 70 atoms. Note it is two-sided: unlike Lipinski it rejects molecules for being too small as well as too large.
• Egan — just two axes, cLogP and TPSA, fitted to passive absorption data. The simplest of the set, and often as predictive as the others.

They overlap heavily. Applying all four is not four times as informative as applying one; it mostly compounds the same information and produces a narrower list for reasons that are hard to articulate afterwards.

Here is the idea that changes how you read all of the above.

Optimisation adds. Over a programme, molecular weight typically rises by 80 to 100 and logP by one to two units, as chemists add groups to pick up contacts. That is not a failure of discipline; it is what happens when you improve potency by making more interactions.

So if your starting point already sits at the Rule of Five limits, your finished drug will be well past them. A lead should therefore be smaller and less lipophilic than the drug you intend to end up with. The usual guide is molecular weight 200 to 350 and cLogP under 3.

This is why a screening hit that passes the Rule of Five comfortably is not necessarily good news: comfortably passing is the minimum, not the target.

Fragment-based discovery starts smaller still: molecular weight ≤ 300, cLogP ≤ 3, donors ≤ 3, acceptors ≤ 3, rotatable bonds ≤ 3. Fragments bind weakly — millimolar is normal — and are grown or linked into leads. Module 22's efficiency metrics were invented for precisely this context, because you cannot compare a fragment and a lead on potency.

A whole class of successful modern drugs sits outside this space deliberately.

• Macrocycles — large, but pre-organised and often able to hide their polar groups internally when crossing a membrane and expose them when binding. This chameleonic behaviour is exactly the intramolecular hydrogen bonding of module 8.
• Peptides and peptidomimetics — many donors and acceptors, usually given by injection.
• Targeted protein degraders — two binding groups joined by a linker, so molecular weight around 800 to 1,000 by construction. They fail every filter here and several are in the clinic.

Ciclosporin is the standing counterexample: molecular weight 1,202, eleven residues, and orally bioavailable. It manages it by macrocyclisation, N-methylation and internal hydrogen bonding.

The lesson is not that the rules are wrong. It is that they describe a well-travelled region of chemical space, and leaving it means giving up the accumulated guidance and needing a specific reason and specific data.

The practical summary of the module.

• Know what the filter was asking. Rule of Five: oral absorption. Rule of three: fragment screening. Applying a fragment rule to a drug candidate is a category error.
• Use them to rank, not to delete. Sorting a list by how badly it violates preserves information; a hard cut destroys it.
• Know when to switch one off. Not an oral drug? The Rule of Five is close to irrelevant. Working on degraders or macrocycles? All of these are.
• Report which count you used, because module 12's ambiguity means the verdict depends on it.
• Never say "it failed Lipinski" as a conclusion. Say which criterion fired, by how much, and why that matters for this project.