Tuberculosis (TB) remains the world’s deadliest infectious disease and one of the hardest to treat. Current therapy involves a cocktail of drugs taken for at least six months, yet nearly one in five patients develop resistance to these first-line treatments. A new study in Cell Systems introduces an AI-powered approach to uncover how TB drugs actually kill bacteria—paving the way for smarter drug combinations that could shorten treatment times and improve cure rates.

The urgent push for stronger therapies
Developing faster, more effective TB regimens is a pressing global health goal. We need a multidrug strategy with three to five new agents that remain effective even against drug-resistant TB, says Bree Aldridge, senior author of the study and professor of molecular biology and microbiology at Tufts University School of Medicine.

Progress has lagged partly because researchers lacked precise tools to determine how TB drugs work and how best to combine them. TB likely has several Achilles’ heels that we could target simultaneously, Aldridge explains. But pinpointing exactly how a drug destroys its bacterial target has been surprisingly difficult.

She likens the process to reconstructing a fight from a crime scene: “You walk into a room and see bruised faces, a toppled chair, and a shattered lamp. You know a fight took place—but not who started it or how it played out.” In the same way, scientists could tell when drugs killed TB bacteria, but not the precise ‘mechanism of death.’

Introducing DECIPHAER
To change that, Aldridge and her colleagues at Tufts University School of Medicine and partner institutions developed DECIPHAER—short for decoding cross-modal information of pharmacologies via autoencoders. This AI-driven system reveals, at a molecular level, how potential TB drugs attack and kill bacteria.

The method builds on the team’s earlier work using high-resolution imaging to capture TB bacteria at the point of death. These so-called morphological profiles serve as crime scene snapshots, showing the structural damage caused by a drug’s mode of action.

“If you treat TB bacteria with a new compound and the cells collapse in the same way they do with other drugs that break down the cell wall, you can infer that the new drug also targets the cell wall,” Aldridge explains.The key advancement with DECIPHAER is its ability to connect visual clues with transcriptional profiles—data revealing which bacterial genes are activated or suppressed during treatment. By training an AI model on both types of information, the system can predict molecular effects using images alone.

Previously, researchers could only guess how TB drugs worked from bacterial shape changes. Now, Aldridge says, DECIPHAER offers precise insights into how drugs damage cells and why bacteria die—revealing, for example, that one candidate drug targets energy production rather than the cell wall.

By predicting molecular effects directly from images, DECIPHAER provides a faster, cheaper alternative to RNA sequencing for testing drug combinations. Aldridge notes her team will keep using it in TB studies and hopes it will speed global drug development, with potential applications beyond TB to other infections and even cancer.

Source: https://www.drugtargetreview.com/news/185159/new-ai-method-maps-how-tuberculosis-drugs-destroy-bacteria/