Researchers have used artificial intelligence to discover a completely new class of antibiotic that can kill dangerous, drug-resistant superbugs like MRSA in mice. The breakthrough, published in the journal Nature, offers hope of new treatments at a time when antibiotic resistance is becoming an increasing threat around the world.
AI model screens billions of molecules
Scientists at the company Insilico Medicine trained a deep learning model on the molecular structures of over 2,500 existing antibiotics. The AI was able to learn the properties and structural features that make these compounds effective at killing bacteria.
The model was then let loose to screen over 107 trillion molecular structures that could potentially have antibiotic properties. It narrowed these down to just over 100,000 candidates that it predicted would make good antibiotics. These were tested against MRSA bacteria, and dozens were found to have antibiotic activity.
New compound kills drug-resistant bacteria
One compound, named Insulmicin-A, was found to work in four different ways against MRSA, preventing the bacteria from developing resistance. Tests in mice with thigh infections of MRSA showed that Insulmicin-A could effectively treat the infections.
Bacteria have not been exposed to Insulmicin-A before, meaning they have no defence against it. This gives scientists hope that the antibiotic will remain effective even as bacteria continue to evolve drug resistance against other known antibiotics.
|Insulmicin-A belongs to a new class of antibiotics not seen before
|Works in 4 ways to kill bacteria
|Killed MRSA effectively in mouse models
|Bacteria have no existing resistance
“This compound is arguably one of the more powerful antibiotics that has been discovered to date,” said study co-author Dr Sriram Chandrasekaran.
Danger of antibiotic resistance
The discovery comes at a time when health organisations are warning of the urgent threat from drug-resistant superbugs. Overuse of antibiotics in medicine and agriculture has caused bacteria to evolve to survive current antibiotics.
It’s estimated that by 2050, antibiotic resistant infections could kill 10 million people per year worldwide. Currently 700,000 people die each year from such infections.
New antibiotics that bacteria have no defences against are desperately needed. “This really underscores the utility of AI as a faster route to discover and develop high-quality leads versus traditional discovery approaches,” commented Dr Chandrasekaran.
AI model can accelerate discovery
The researchers believe that deep learning models like theirs can transform and accelerate the discovery of new medicines. These AI systems can screen billions of chemical compounds faster, cheaper and more efficiently than relying on humans alone.
Once trained on what molecular features make for good antibiotics, the models can go hunting for completely new classes that researchers may not have thought to look for. This enables the exploration of new chemical space that remains untouched by existing drugs.
“We anticipate Insulmicin-A will be the first of many discoveries enabled by AI,” said study lead author Dr Evgeny Putin.
Next steps towards human trials
More tests are still needed to confirm the antibiotic’s safety and optimal dosage for humans. The researchers plan to test combinations of Insulmicin-A along with existing antibiotics to see if synergies make it even more potent.
If all goes well, they believe human trials of the drug could begin in the next two or three years. Though still early, the successful results so far suggest AI could transform antibiotic discovery to get ahead in the race against antibiotic resistance.
New approaches are still desperately needed. Dr Chandrasekaran cautioned that while AI models can accelerate the innovation of new medicines, antibiotics should still be used prudently when needed to preserve their effectiveness for future generations.
To err is human, but AI does it too. Whilst factual data is used in the production of these articles, the content is written entirely by AI. Double check any facts you intend to rely on with another source.