Modeling membrane permeation to optimize pharmacokinetics

There are four research areas in the Department of Pharmaceutical Chemistry. Modeling membrane permeation to optimize pharmacokinetics is a research challenge within computational chemistry and biology.

The challenge

An effective drug molecule must not only selectively bind and therapeutically modulate its protein target; it also must have appropriate in vivo absorption, distribution, metabolism, and excretion characteristics (ADME). For drug targets inside cells, drugs and chemical biology tools must be able to enter the cell to be effective. For drugs taken by mouth, its physicochemical properties must also allow it to reach its target by passing through membranes in the gut (oral bioavailability). In addition, less than five percent of small molecule drugs are able to pass through the blood-brain barrier in order to treat, for example, neurodegenerative disease, psychiatric disorders, or epilepsy.

Using computer models to predict which potential drug compounds are best able to reach their targets in the body, as well as how to modify them to improve their ability to permeate key membranes, would be a faster and more cost-effective way to develop new and better drugs.

Examples of our research and methods include

Department research develops computer programs that model the biophysics of membrane permeation to predict and improve the ability of potential drug compounds to pass through bodily barriers and enter cells.

The biophysical approach taken here is distinct from methods primarily now in use for drug design. These methods seek to predict a molecule’s membrane permeability by statistically comparing its features to chemically similar ones with experimentally determined permeability. Such quantitative structure-permeability relationship (QSPR) methods can be limited in applicability to novel compounds and may not provide the mechanistic insight needed to improve a given drug lead’s permeability.

Department scientists seek more broadly applicable and systematically improvable modeling that emphasizes the role of molecules’ conformational flexibility (performing extensive conformational sampling) and their ability to form internal hydrogen bonds. These bonds increase the molecules’ greasy water-repelling quality (hydrophobicity), allowing it to more readily pass through also-hydrophobic cell membranes.

Research here also develops programs to predict the binding of molecules as substrates by membrane-associated efflux transporter proteins such as P-glycoproteins (P-gp), which prevent exogenous substances, including drugs, from crossing the intestinal lining or entering other sensitive organs, including the brain. Beyond affecting drugs’ distribution and elimination, the efflux proteins’ overexpression in cancer cells has been linked with multi-drug resistance in tumors. Existing experimental assays are expensive, time-consuming, and limited in their accuracy for certain substrates, while quantitative structure-activity relationship (QSAR) modeling (i.e. comparing structural features to those of existing drug molecules less subject to efflux) is complicated by the chemical diversity of P-gp substrates.

Cyclic peptides are being used as a challenging system for studying membrane permeation because:


A schematic diagram of the ring-link cyclic peptide structure of the cyclosporine molecule. Cyclosporine is an immunosuppressant drug used to prevent organ rejection in transplant patients.

  • Their larger size and relative complexity could provide scaffolds for oral drugs targeting protein-protein interactions as well as proteins without well-defined small molecule binding sites.
  • Although more than 100 macrocyclic drugs are in use, their greater molecular weights (i.e., 800 daltons vs. less than 500 Da for most small molecule drugs) and complexity (e.g., number of polar groups) are not traditionally drug-like. Thus they require new ways to understand, predict, and improve their pharmacokinetics, including gut absorption.
  • They can be readily synthesized with well-defined variations in factors, such as the spatial arrangement of key atoms (stereochemistry), rigidity, hydrophobicity, etc., and so predictive computational models can be promptly followed and iteratively improved by experimental testing.

Examples of such computational chemistry and biology research by department scientists include

Predicting blood-brain barrier penetration

Department scientists developed a new computational approach combining mechanistic modeling of passive membrane permeation and active efflux by P-glycoprotein to better predict which compounds can penetrate the blood-brain barrier. This included demonstrating that the dual-parameter approach is better than either alone at predicting experimentally derived high vs. low efflux ratios and the ability to permeate the barrier. It was also shown that the approach could be used to rank-order similar compounds for permeability and provide guidance into modifying molecules to improve their ability to enter the brain.

Calculated passive permeability and P-glycoprotein efflux for compounds experimentally assayed for their ability (red) or inability (blue) to penetrate a monolayer of canine kidney cells modified to overexpress P-glycoprotein (an in vitro model for predicting blood-brain barrier penetration of potential drug molecules). The dotted colored lines correspond to computed P-gp efflux ratios.

Researchers here developed a new algorithm for predicting the binding specificity—and thus likely substrates—of the P-glycoprotein (P-gp) efflux transporter protein, using flexible receptor docking as opposed to QSAR. The approach was driven by structural studies of mouse P-gp bound with drug-like substrates that revealed a large, hydrophobic binding pocket without clearly defined sub-sites. This is supportive of the induced-fit binding model in which an enzyme’s active site substantially changes conformation in accommodating diverse ligands. The computational method for predicting P-gp substrates was tested and confirmed against experimental assay data, as well as with a series of peptidic cysteine protease inhibitors that are being developed as anti-parasitic drug leads.

Improving membrane permeability of peptides

  • Department scientists predicted that the ability to form intramolecular hydrogen bonds could improve membrane permeation, using a molecular modeling approach. This computational finding was confirmed experimentally by the synthesis of pairs of very similar molecules, differing primarily in their ability to form hydrogen bonds. Thus researchers here demonstrated a way to improve membrane permeability while retaining or improving other favorable drug-like properties.
  • Department scientists and collaborators used computer modeling to predict the effect of specific configurations of backbone N-methylation on the membrane permeability and thus the oral bioavailability of cyclic peptides. Their model worked from the hypothesis, drawn from naturally bioactive cyclic peptides, that such N-methylation at specific locations alters conformation to stabilize intramolecular hydrogen bonding in the peptides’ membrane-associated state, improving their hydrophobicity and thus their permeability. This conformational hypothesis was experimentally tested by synthesizing N-methylated cyclic hexapeptide scaffolds that had in vitro permeability comparable to small molecule drugs.


Schematic representation of the conformational hypothesis of membrane permeability applied to N-methyl cyclic peptides