druglikeness

Druglikeness is a loose, holistic judgment of whether a molecule 'looks like' the kind of compound that tends to become a successful oral drug. It is less a single number than a family resemblance, learned from the shared traits of medicines already on pharmacy shelves: reasonable size, balanced lipophilicity, modest polarity, and chemistry that is stable and not flagrantly reactive.

Many tools try to capture it. Simple filters such as Lipinski's rule of five and Veber's criteria set soft limits on weight, logP, hydrogen bonds, polar surface area, and rotatable bonds. Composite scores such as the quantitative estimate of druglikeness (QED) blend several properties into one value, and various rule sets also weed out unstable groups or known nuisance substructures.

It is a probabilistic prior, not a verdict. Druglikeness filters trim a screening library toward sensible chemistry early on, but they can wrongly reject good molecules and rubber-stamp bad ones, and they explicitly do not judge potency, selectivity, or safety. Whole classes of real drugs, from antibiotics to macrocycles, sit outside the conventional druglike box.