Lipinski's Rule of Five

The first criterion limits molecular size. High molecular weight compounds (>500 Da) struggle to passively diffuse through the tightly packed lipid bilayers of intestinal epithelial cells. Larger molecules also tend to have poorer aqueous solubility, which directly limits the concentration gradient driving passive absorption. This cutoff is not absolute—many successful macrocyclic drugs and natural products exceed 500 Da—but the probability of oral activity declines sharply above this threshold.

This criterion governs lipophilicity, the compound's affinity for lipid environments over aqueous phases. A high LogP (>5) indicates extreme hydrophobicity, causing the molecule to sequester irreversibly in membrane lipids or bind non-specifically to hydrophobic protein pockets. Such compounds also suffer from poor aqueous solubility and rapid metabolic clearance. Conversely, excessively hydrophilic molecules (LogP < 0) cannot partition into the lipophilic core of the cell membrane.

The sum of all O–H and N–H bonds in the molecule. Hydrogen bond donors represent the compound's capacity to donate protons to surrounding water molecules. Each HBD requires substantial solvation energy to desolvate before the molecule can enter the hydrophobic membrane interior. Excessive HBDs (>5) impose a prohibitive energetic penalty on passive membrane permeation. This is typically expressed as the sum of all hydroxyl and amine groups.

The sum of all nitrogen and oxygen atoms in the molecule. HBAs represent lone-pair electrons available to accept hydrogen bonds from surrounding water. Like HBDs, each HBA must shed its hydration shell before the molecule can partition into the membrane. A high HBA count (>10) correlates strongly with poor permeability and is a common feature of compounds designed to violate the rule, such as highly polar peptidomimetics.

The logarithm of a compound's partition coefficient between octanol and water, serving as the key quantitative measure of molecular lipophilicity. It directly corresponds to the second rule (LogP ≤ 5). High LogP values indicate poor aqueous solubility and extensive plasma protein binding, while excessively low values prevent passive membrane permeation. Modern in silico models predict LogP using atomic contribution methods or graph neural networks trained on experimental shake-flask data.

The fraction of an orally administered dose that reaches systemic circulation unchanged. It is the composite pharmacokinetic parameter that the Rule of Five attempts to optimize. Bioavailability is governed by a complex interplay of aqueous solubility, intestinal permeability, and first-pass metabolism in the liver. While Lipinski's rules provide a rapid filter, modern physiologically based pharmacokinetic (PBPK) models offer a more mechanistic prediction of the fraction absorbed.

The sum of the atomic weights of all atoms in a molecule, forming the first and most intuitive rule (≤ 500 Daltons). Molecular weight correlates strongly with membrane permeability; larger molecules diffuse more slowly across lipid bilayers. It also serves as a proxy for molecular complexity. In beyond-Rule-of-5 (bRo5) space, macrocyclic peptides and PROTACs often exceed this limit, requiring active transport mechanisms or endocytosis for cellular uptake.

The third and fourth rules limit hydrogen bond donors (sum of OH and NH groups ≤ 5) and acceptors (sum of N and O atoms ≤ 10). These constraints reflect the energetic cost of desolvating polar groups to cross a lipid membrane. Each hydrogen bond must be broken to move from an aqueous environment into the hydrophobic core of a bilayer. Intramolecular hydrogen bonding can mask polarity, allowing some molecules to violate these rules while maintaining good permeability.

The chemical space beyond Lipinski's physicochemical boundaries, populated by macrocycles, PROTACs, and cyclic peptides. These compounds often violate multiple rules simultaneously yet achieve oral bioavailability through chameleonicity—the ability to adapt their conformation to the environment. In polar media, they expose hydrogen bonds; in non-polar media, they form intramolecular hydrogen bonds to shield polarity. This dynamic behavior challenges static rule-based filters.

The computational estimation of a drug candidate's Absorption, Distribution, Metabolism, Excretion, and Toxicity properties. While Lipinski's rules focus narrowly on oral absorption, modern ADMET models predict a comprehensive pharmacokinetic profile using multi-task neural networks. These models integrate heterogeneous data—from Caco-2 permeability assays to hERG cardiotoxicity screens—to provide a holistic early-stage safety assessment before costly synthesis begins.