Tuberculosis claimed 1.23 million lives in 2024, remaining one of the world’s deadliest infectious diseases due to the bacterium’s highly selective outer shell, the mycomembrane. Researchers at the University of Massachusetts Amherst have now developed a paired screening and machine-learning approach to identify which drug compounds can successfully penetrate this barrier, according to findings published in Nature Microbiology.
Overcoming the Mycomembrane Barrier
The primary hurdle in treating Mycobacterium tuberculosis (Mtb) is its physical structure. Mtb is shielded by a mycomembrane that acts as a gatekeeper, blocking many drugs from reaching their targets.
"Mtb is unique," said Sloan Siegrist, associate professor of microbiology at UMass Amherst and one of the study’s senior authors. "Not only does it have two membranes that protect the cell from antimicrobial chemical compounds that we might use to kill it, its outer membrane is unlike any other biological barrier out there."
Traditional drug discovery often fails because candidate compounds that look promising on paper cannot actually enter the bacterial cell. By the time researchers realize a drug cannot penetrate the membrane, they have already invested significant time and resources into a dead-end candidate.
The PAC-MAN Screening Technique
To address this, researchers utilized a method called Peptidoglycan Accessibility Click-Mediated AssessmeNt (PAC-MAN). Instead of testing whether a drug kills the bacteria, PAC-MAN measures whether a molecule can successfully cross the mycomembrane.
In the study, the team screened more than 1,500 azide-tagged small molecules. The results revealed that specific chemical structures, particularly aromatic nitrogen-containing heterocycles like indole, imidazole, and pyrazole, were more likely to pass through the outer layer. Conversely, structures such as cyclopentane and cyclohexane were associated with lower permeability.
The mycomembrane is so selective that standard chemical traits such as lipophilicity and polar surface area showed weak or inconsistent relationships across the full dataset. The UMass Amherst team found that the effect of a property depended strongly on the rest of the molecule, meaning a feature that helps one class of compounds cross the mycomembrane may do little, or even the opposite, in another.
Predicting Permeability with MycoPermeNet
Building on the PAC-MAN data, the research team developed a machine learning model named MycoPermeNet. Led in part by Anna Green, an assistant professor in UMass Amherst’s Manning College of Information and Computer Sciences, the model was trained to predict whether a compound would cross the mycomembrane based on its chemical structure.
"Small molecules can be particularly difficult to analyze computationally," Green said. "Because they come in all different sizes with a wide range of molecular connections, you can’t describe them with a single measurement, by weight, say, or size."
The model proved effective at identifying molecular features that facilitate entry. When the team tested these predictions by swapping specific ring structures in test molecules—such as replacing phenylalanine side chains with tryptophan—they observed the predicted increase in membrane permeation.
Future Directions for Tuberculosis Treatment
While this new approach does not deliver a finished tuberculosis drug, it provides a faster, more targeted way to prioritize candidates. By filtering out non-permeable molecules early, researchers can focus their efforts on compounds that have a higher probability of reaching their target inside the cell.
However, the authors noted that permeability is only one factor. Even if a drug enters the cell, it must still survive processes like metabolism and target binding. In some cases, improvements in membrane permeability did not track cleanly with growth inhibition, indicating that the mycomembrane is not always the main barrier limiting a drug.
Despite this, the integration of PAC-MAN and MycoPermeNet represents a shift toward more intelligent drug design, potentially reducing the time required to move promising candidates from the computer screen to clinical development.
Frequently Asked Questions
Why is tuberculosis so difficult to treat with drugs?
The Mycobacterium tuberculosis bacterium is protected by an unusual, highly selective outer layer called the mycomembrane. This barrier blocks many drugs from entering the cell, rendering them ineffective before they can reach their target.
What is the PAC-MAN method?
PAC-MAN (Peptidoglycan Accessibility Click-Mediated AssessmeNt) is a screening technique that measures whether a molecule can penetrate the mycomembrane. It focuses on the physical entry of the drug rather than the final killing effect.
How does MycoPermeNet assist in drug discovery?
MycoPermeNet is a machine learning model that predicts whether a drug compound will be able to cross the Mtb mycomembrane based on its chemical structure. This allows researchers to prioritize only the most promising molecules, saving time and resources.
Does membrane permeability guarantee a drug will work?
No. While entry is essential, a drug must also successfully bind to its target and avoid being broken down or pushed out of the cell by the bacteria. The researchers found that permeability is a necessary, but not always sufficient, condition for success.
Are you interested in the intersection of artificial intelligence and infectious disease research? Subscribe to our newsletter for the latest updates on medical breakthroughs and clinical trials.
