Tuberculosis remains the world’s deadliest single-agent infection, claiming 1.23 million lives in 2024. A major obstacle in developing effective treatments is Mycobacterium tuberculosis, whose unique mycomembrane effectively blocks most antimicrobial compounds. To overcome this barrier, a multidisciplinary research team led by the University of Massachusetts Amherst and the University of Virginia has engineered a combined high-throughput screening and artificial intelligence pipeline that dramatically accelerates the identification of potent tuberculosis drugs. The findings were recently published in Nature Microbiology. Historically, screening chemical compounds against Mtb required testing each variant individually, a prohibitively slow process given the vast chemical space. To address this, the team previously developed Peptidoglycan Accessibility Click-Mediated AssessmeNt, or PAC-MAN, a laboratory technique introduced in 2023 that enables parallel testing of multiple compounds simultaneously. While PAC-MAN significantly improved experimental throughput, the researchers sought to extend its utility beyond direct measurement. By integrating computational biology and machine learning, the team created MycoPermeNet, a specialized neural network trained on PAC-MAN screening data. The model analyzes molecular structure alone to predict a compound’s ability to permeate the mycomembrane, identifying specific physical properties and substructural features that facilitate cellular entry. This AI-driven approach eliminates the need to manually evaluate every candidate, allowing researchers to rapidly filter millions of virtual molecules for viable leads. The synergy between PAC-MAN and MycoPermeNet has yielded critical insights into Mtb drug penetration. The AI model revealed that the same molecular attributes governing membrane permeability strongly correlate with a compound’s ability to kill the bacterium. Consequently, the platform not only predicts which chemicals can bypass the mycomembrane but also flags those with high therapeutic potential. This dual validation mechanism streamlines the early stages of antibiotic development, reducing reliance on labor-intensive trial-and-error screening. Sloan Siegrist, associate professor of microbiology at UMass Amherst, and Marcos Pires, professor of chemistry at the University of Virginia, co-developed the foundational screening methodology. Anna Green, assistant professor at UMass Amherst’s Manning College of Information and Computer Sciences, led the computational modeling effort. The collaboration bridges experimental microbiology, synthetic chemistry, and artificial intelligence, establishing a reproducible framework for tackling other pathogens protected by resilient cellular barriers. By transforming how researchers evaluate chemical uptake, the PAC-MAN and MycoPermeNet pipeline offers a scalable pathway to combat drug-resistant tuberculosis. The integration of high-throughput experimentation with machine learning prediction marks a significant advancement in antimicrobial discovery, potentially shortening development timelines and improving global public health outcomes against a persistent infectious disease threat.