PAMPA

PAMPA utilizes a lipid-infused artificial membrane—typically a mixture of phospholipids dissolved in an organic solvent like dodecane—coated onto a porous filter support. This creates a simplified, reproducible model of the intestinal epithelial bilayer. Unlike cell-based assays (e.g., Caco-2), it lacks active transporters and paracellular pores, isolating the passive transcellular diffusion mechanism. The lipid composition can be tuned; for example, a porcine brain lipid extract is used to model the Blood-Brain Barrier (BBB-PAMPA).

To maintain sink conditions and mimic in vivo absorption, a pH gradient and a scavenger are often employed. The donor compartment is typically buffered at pH 5.5–6.5, while the acceptor compartment is at pH 7.4. A surfactant or serum protein in the acceptor well binds permeated molecules, preventing back-diffusion. This 'double-sink' approach creates a unidirectional flux, ensuring the effective permeability (P_e) measurement reflects true absorption potential rather than an equilibrium state.

The effective permeability (P_e) is calculated from the rate of compound appearance in the acceptor compartment, typically expressed in cm/s. Results are benchmarked against known standards:

• High Permeability: P_e > 10 × 10⁻⁶ cm/s (e.g., verapamil)
• Low Permeability: P_e < 1 × 10⁻⁶ cm/s (e.g., ranitidine) This binary classification directly informs the Biopharmaceutics Classification System (BCS) and correlates strongly with human oral absorption fraction.

A critical artifact in PAMPA is the Unstirred Water Layer (UWL), a stagnant aqueous boundary adjacent to the membrane. For highly permeable lipophilic compounds, diffusion across the UWL becomes the rate-limiting step, causing an underestimation of true membrane permeability. Advanced protocols apply a pKa^{flux} correction or use vigorous stirring to minimize UWL thickness. In silico models trained on PAMPA data must account for this, often by incorporating Abraham solvation parameters to deconvolute UWL resistance from membrane resistance.

PAMPA data serves as a gold-standard training endpoint for Quantitative Structure-Property Relationship (QSPR) models. Key molecular descriptors used to predict PAMPA permeability include:

• LogP/LogD: Partition coefficient as a measure of lipophilicity
• Topological Polar Surface Area (TPSA): A proxy for hydrogen bonding capacity
• Molecular Weight: Correlates inversely with diffusion coefficient Graph Neural Networks (GNNs) trained on PAMPA datasets can generalize beyond the Lipinski rule-of-five space, predicting permeability for macrocycles and PROTACs.

While both measure permeability, they provide distinct information:

• PAMPA: Isolates passive transcellular transport. High-throughput, low cost, and highly reproducible.
• Caco-2: A cell-based assay expressing active transporters (e.g., P-gp efflux) and tight junctions. It captures active uptake, efflux, and paracellular transport. A high PAMPA permeability but low Caco-2 permeability strongly suggests the compound is an efflux transporter substrate, a key liability flagged in drug discovery.