PAMPA Assay for Measuring Passive Drug Permeability

The PAMPA assay, short for Parallel Artificial Membrane Permeability Assay, is a lab-based screening tool used primarily in drug discovery to measure how easily a compound passes through a synthetic membrane by passive diffusion. It works by placing a drug candidate in a donor well, separating it from a receiver well with a thin artificial lipid membrane, and then measuring how much compound crosses over a set period. The technique has become one of the most widely used early-stage permeability screens in the pharmaceutical industry because it is fast, inexpensive, and highly amenable to automation, capable of processing thousands of compounds in a single run.

What PAMPA Actually Measures

At its core, PAMPA measures passive transcellular permeability. That means it tells you whether a molecule can dissolve into and cross a lipid barrier on its own, without help from transporter proteins or other biological machinery. The assay uses a 96-well plate format: the bottom wells hold the donor solution containing the test compound, the top wells hold the acceptor solution, and sandwiched between them is a filter support coated with an artificial lipid mixture. After an incubation period, typically four to sixteen hours, the concentration of compound in the acceptor compartment is measured, usually by UV spectroscopy or liquid chromatography paired with mass spectrometry. LC/MS detection is particularly useful for compounds that lack a strong UV signal or have low water solubility, since solubilizing agents that would interfere with UV readings can be used without compromising the measurement.¹

The result is an apparent permeability coefficient, often written as P_app, reported in units of centimeters per second. Higher values mean the compound crosses the membrane more readily. Because the membrane is entirely synthetic, no cells are involved and no active transport or efflux mechanisms exist in the system. This is both its chief advantage and its chief limitation: the assay cleanly isolates passive diffusion, but it tells you nothing about whether a drug would be pumped out of cells or shuttled across by carrier proteins.²

How PAMPA Compares to Caco-2 Cell Models

The other major in vitro permeability method you will hear about is the Caco-2 assay, which uses a monolayer of human colon cancer cells grown on a filter support. These cells form tight junctions and express transporter proteins, so they capture paracellular transport (movement between cells), carrier-mediated uptake, and efflux. The tradeoff is that Caco-2 cultures take roughly three weeks to grow, require careful maintenance, and are significantly more expensive per plate. PAMPA, by contrast, needs no cell culture at all and can be run the same day the plates are prepared.³

Despite these differences, the two methods agree more often than not. Comparisons using sets of marketed drugs show at least 85% concordance between PAMPA and Caco-2 permeability rankings.⁴ In broader compound sets, the two methods correlate with R² values ranging from about 0.76 to 0.92, indicating that passive diffusion dominates the permeation of most drug-like molecules.⁵

Where the numbers diverge is where the interesting biology lives. When a compound shows high PAMPA permeability but low Caco-2 permeability, efflux transporters are likely pumping it back out of the cell. When Caco-2 permeability is higher than PAMPA, active uptake or paracellular transport is probably helping it across. Running both assays in tandem therefore gives medicinal chemists a way to flag compounds that may rely on or be penalized by active transport mechanisms.⁶ In practice, many pharmaceutical companies use PAMPA as the first-pass screen to triage thousands of compounds quickly and then send promising hits through Caco-2 for a more biologically complete picture.

Variants Designed for Different Biological Barriers

Not every membrane in the body behaves like the gut lining. The blood-brain barrier is far more restrictive, and the skin’s outermost layer, the stratum corneum, has its own lipid architecture. Researchers have modified the basic PAMPA setup to mimic these barriers by changing the lipid composition of the artificial membrane.

PAMPA-BBB uses a lipid mixture formulated to approximate the tight, cholesterol-rich membranes of brain capillary endothelial cells. Developed using a training set of structurally diverse commercial drugs, the assay predicts passive blood-brain barrier penetration with high accuracy and retains the throughput and reproducibility advantages of standard PAMPA.⁷ Skin PAMPA, meanwhile, coats the filter with a blend of free fatty acid, cholesterol, and synthetic ceramide analogs to simulate the stratum corneum’s barrier function.⁸ Compatibility testing has shown that certain formulation additives and acceptor media components, including DMSO and ethanol at concentrations up to 50%, do not compromise the skin PAMPA membrane’s integrity during standard incubation times.⁹

A further refinement is the double-sink PAMPA, which uses a pH gradient between the donor and receiver compartments (commonly pH 6.5 and 7.4, respectively) along with a surfactant in the receiver solution to act as a lipophilic sink. The pH gradient mimics the shift from the slightly acidic environment at the intestinal wall surface to the more neutral pH of the blood, and the sink condition prevents back-diffusion, improving the dynamic range of the measurement.¹⁰ pH-permeability profiling done with these gradient setups has shown good agreement with human intestinal absorption data, confirming that pH strongly influences how efficiently drugs cross the gut wall.¹¹

The Unstirred Water Layer Problem

One of the trickiest technical issues in PAMPA, and one that has driven years of methodological work, involves a layer of stagnant liquid that sits on each side of the membrane. In a standard, unstirred plate setup, this unstirred water layer can be 1,500 to 4,000 micrometers thick.¹² Inside the human intestine, the equivalent layer is estimated at only about 30 to 100 micrometers.¹³

Why does this matter? For highly lipophilic, fast-permeating molecules, the artificial membrane itself is not the bottleneck. Instead, the rate-limiting step becomes how quickly the compound can diffuse through the stagnant water on either side of the membrane. When that happens, the assay is effectively measuring the permeability of water, not the membrane, and the resulting numbers underestimate the compound’s true membrane permeability. Without correction for this effect, the in vitro permeability coefficient of lipophilic molecules reflects only the property of water.¹⁴

Gently rocking the plate on a standard shaker helps, but it cannot thin the water layer enough to reach physiologically relevant values. Individual-well magnetic stirring at speeds above 110 rpm has been demonstrated to bring the unstirred water layer down to the in vivo range, and at 622 rpm the layer has been thinned to as little as 13 micrometers, a feat not previously achieved in plate-based permeability assays.¹⁵ Labs that do not use this level of stirring need to apply mathematical corrections to their data, which in turn depend on accurate estimates of the filter porosity. If the effective porosity of the filter is wrong, both the intrinsic membrane permeability and the unstirred water layer thickness can be significantly miscalculated.¹⁶

Membrane Retention of Lipophilic Compounds

A related but distinct challenge arises with very lipophilic drug candidates. These molecules tend to dissolve into the artificial membrane and stay there rather than crossing all the way into the acceptor compartment. If you only measure what shows up in the donor and acceptor wells at the end of the experiment, you may underestimate how much compound actually left the donor side, leading to skewed permeability values. In one study of marketed drugs, chlorpromazine, a famously greasy molecule, showed 64% membrane retention in PAMPA, while most other standard compounds were retained at up to about 26%.¹⁷

The problem has become more pressing as modern drug discovery pipelines increasingly yield candidates with poor water solubility and high lipophilicity. Drugs with higher lipophilicity undergo more membrane retention, which is consistent across different assay formats.¹⁸ Accounting for the fraction trapped in the membrane is now considered essential for getting accurate permeability numbers from any lipid-based in vitro system. Some newer assay designs measure the membrane compartment directly to quantify retention alongside transport.

Predicting Human Absorption From Plate-Based Data

The ultimate goal of running a PAMPA experiment is rarely the permeability number itself; it is whether the compound will be absorbed in a living person. Several groups have shown that optimized PAMPA conditions correlate well with the human intestinal absorption fraction. One study found that using a 2% phosphatidylcholine membrane with a pH 6.5/7.4 donor/acceptor gradient produced correlations of r ≥ 0.90 with the fraction absorbed in humans and distinguished high-absorption from low-absorption compounds with no false positives or false negatives in the test set.¹⁹ The same setup simultaneously predicted plasma protein binding, though it did not correlate with volume of distribution, a pharmacokinetic parameter driven by tissue partitioning rather than membrane crossing.

Hexadecane-membrane PAMPA has also been adapted to predict plasma protein binding by incorporating human serum albumin and alpha-1-acid glycoprotein into the receiver compartment.²⁰ These dual-purpose variants are attractive in early discovery because they extract two pieces of pharmacokinetic information from one plate, saving time and compound.

Computational Models Trained on PAMPA Data

Running PAMPA is cheap relative to cell-based assays, but even cheap assays become expensive when you need to screen millions of virtual compounds. This has driven a substantial effort to build in silico models that predict PAMPA permeability from molecular structure alone. A large-scale effort using a dataset of over 5,400 structurally unique molecules measured by a single lab built a support vector regression model that achieved an area under the ROC curve of 0.90 when predicting held-out compounds, with roughly 84% of predictions falling within 0.2 log units of the experimental value.²¹

Other groups have explored combining classical regression approaches, which are easier to interpret mechanistically, with machine learning methods like neural networks and support vector machines, which tend to predict better but behave as black boxes.²² One strategy pairs a partial least squares model (useful for understanding which molecular features drive permeability) with a hierarchical support vector regression model (useful for accurate prediction), letting researchers get both insight and predictive power from the same dataset.²³ As these computational tools mature, they increasingly serve as the first filter in virtual screening campaigns, with physical PAMPA experiments reserved for validating the top-ranked candidates.

Testing Excipients and Formulation Effects

PAMPA is not limited to ranking drug molecules. It has also found a role in formulation science, where researchers want to know whether the inactive ingredients in a tablet or capsule affect how well the active drug crosses membranes. A 2024 comparison of PAMPA, Permeapad, and Caco-2 models tested 14 common pharmaceutical excipients against drugs spanning different solubility and permeability classes. Excipients like hydroxypropyl cellulose and povidone K30 increased the permeability of poorly permeable drugs in the PAMPA model, while croscarmellose sodium and sodium lauryl sulfate decreased the permeability of highly permeable drugs in a concentration-dependent manner.²⁴ These findings have practical consequences for formulators: an excipient chosen to improve tablet disintegration could inadvertently reduce how much drug gets absorbed.

Biological additives have been explored too. In one study, probiotic bacteria significantly increased the membrane permeability of the diabetes drug gliclazide across the PAMPA membrane at multiple pH values, while bile acids decreased it, with deoxycholic acid having a stronger inhibitory effect than cholic acid. The probiotic effect was partly offset at neutral pH by apparent bacterial metabolism of the drug, reducing the total amount available for transport.²⁵ Studies like these use PAMPA not to rank drug candidates but to understand how the formulation and biological environment modulate absorption of an already-approved drug.

Applications Beyond Pharmaceuticals

While drug discovery accounts for the vast majority of PAMPA use, the assay has found a second life in environmental and agricultural chemistry. The logic is straightforward: if the assay can tell you how well a molecule crosses a biological membrane, it can also help predict how readily a pesticide accumulates in fish tissue or how a systemic agrochemical distributes inside a plant.

Researchers have used PAMPA to measure a membrane partition index for organophosphorus pesticides and found that it correlated with bioconcentration factors in fish better than the traditional octanol-water partition coefficient. The results suggest that PAMPA captures something about how molecules interact with real lipid membranes that a simple shake-flask partitioning experiment misses.²⁶ Because the same PAMPA measurement can estimate both human oral absorption and aquatic bioconcentration, the assay offers an efficient way to assess multiple exposure routes for environmental chemicals simultaneously.

In agriculture, PAMPA run under a pH gradient has provided experimental evidence for ion trapping, where acidic or basic agrochemicals accumulate preferentially in plant compartments of opposing pH. This mechanism governs whether a systemic compound ends up in the phloem, the xylem, or trapped inside cells, and understanding it helps chemists design molecules that reach the right tissue within the plant.²⁷ PAMPA’s QSAR models for hydrophobic compounds, developed originally for drug screening, have been extended to pesticides with comparable predictive accuracy.²⁸

Emerging Biomimetic Alternatives

PAMPA is not the only artificial membrane game in town anymore. A newer tool called Permeapad uses a phospholipid-based physical patch rather than a lipid-coated filter. Microscopic examination reveals that the barrier forms phospholipid vesicles and myelin-like structures when exposed to aqueous solutions, creating a more structurally complex interface than the simple lipid layer in traditional PAMPA. Early comparisons showed good correlation with both PAMPA and Caco-2 permeability values for compounds spanning a wide permeability range.²⁹ The technology has since been scaled to a 96-well plate format with correlations of R² = 0.89 against PAMPA and R² = 0.70 against Caco-2.³⁰

Whether barriers like Permeapad will eventually displace PAMPA in routine screening remains to be seen. PAMPA benefits from decades of accumulated reference data, well-characterized lipid formulations for specific biological barriers, extensive computational model coverage, and deep familiarity across the pharmaceutical industry. Any replacement has to match not only the assay’s performance but its enormous installed base of historical data. For the foreseeable future, PAMPA will likely remain the default first-pass permeability screen, with newer technologies carving out roles in situations where its specific limitations, especially around lipophilic compounds and the absence of active transport, matter most for the question at hand.