MIT researchers have used artificial intelligence (AI) to create previously unexisting molecules to develop promising antibiotic candidates for MRSA, a superbug that killed millions.
MRSA, or methicillin-resistant Staphylococcus aureus, is caused by staphylococcus that has become resistant to most antibiotics, according to the Centers for Disease Control. Harvard Medical School found out, “MRSA has proven to be particularly skilled at avoiding grasping antibiotics and becoming truly dangerous superbugs.”
According to the National Institutes of Health (NIH), the common locations for contracting for MRSAs are hospitals, long-term care facilities and communities.
Traditional methods of discovering new antibiotics mean that scientists look at existing chemical databases of molecules discovered in natural or synthetically made compounds and test them against bacteria.
However, these resources are primarily known molecule variations, and discovering new classes of drugs becomes difficult, according to a paper by MIT researchers published in Journal Cell.
Using Gen AI, the MIT team envisioned an entirely new chemical structure that has never been created before. The AI models they used learned patterns from millions of known molecules and remixed them.
“We can use the generator AI to create antibiotics that were not yet present to come up with molecules, and overcome existing resistance mechanisms, which can overcome existing resistance mechanisms.
According to PYMNTS Intelligence Data, the healthcare industry is expected to reach $22 billion by 2032. The healthcare industry is leveraging the possibilities of AI.
Reinventing drug discovery
AI is reinventing drug discovery. According to the NIH, one of the main challenges of drug discovery is the “extensive chemical space” that must be investigated to identify potential drug candidates.
“Traditional methods for screening large compound libraries are labor intensive, time-consuming and often lead to limited number of hits,” the NIH said in its report. “AI-driven virtual screening approaches utilize machine learning algorithms that can rapidly sift through vast data sets of compounds and predict biological activity against specific drug targets.”
According to the NIH, using AI could reduce the time between identifying promising drugs and market release. This usually takes 15 years.
At MIT, researchers could use AI to sift over 209 million molecules to 22 to narrow them down to 22, working and realistically synthesize them. Six of them showed promises, and one candidate DN1 was particularly effective in clearing MRSA from mice.
MIT researchers use the same procedure to find promising antibiotics and fight multidrug resistant Neisseria goNorrhoeae, a bacteria that leads to go disease, a common sexually transmitted disease. Of the 45 million people in the database, the team narrowed it down to two that can be synthesized using AI. Only one was effective: NG1.
The authors noted that scientists allow scientists to tap on the vast potential of molecules that could generate classes of new drugs. As antibiotic resistance causes around 5 million deaths annually worldwide, this approach may help large pharmaceutical companies revive long-negleeded areas. They said that between 1980 and 2003, only five antibacterial agents were developed by the top 15 drug makers.
The team is part of MIT's antibiotic-AI project created to fill the gaps left by major pharmaceutical companies. Their partner in the study, nonprofit Phare Bio, is further modifying DN1 and NG1 to prepare for further testing. Phare Bio is also part of MIT's antibiotics program.
Asked when these antibiotics will be prepared for the public, Collins said: “We expect it will take several years for AI-designed molecules to be approved for public use. We are working with nonprofit Phare Bio to optimize and advance the most promising molecules clinically.”
read more:
Can it cure cancer? Experts say the answer is complicated
Stanford AI Agents'Lab is finding promising Covid-19 drug leads in just a few days
Sandboxaq launches datasets to train AI models for drug discovery
