When you get a scrape, you use an antibiotic ointment. When you get a chest infection, your doctor prescribes antibiotics.
Ever since the discovery of penicillin in 1929, scientists have created various antibiotics to help treat diseases like pneumonia, strep throat, and scarlet fever. However, antibiotics are costly and time-consuming to make.
Scientists at the Massachusetts Institute of Technology (MIT) are attempting to solve these problems by using artificial intelligence (AI) to create antibiotics. Let’s learn more!
Why Do We Need New Antibiotics?
The more people use antibiotics, the less effective these medicines typically become.
This is called antibiotic resistance and occurs when an antibiotic fails to kill all target pathogens (disease-causing bacteria and other tiny organisms), leaving only the immune ones. These surviving pathogens then reproduce, passing on their antibiotic immunity and creating a new generation of stronger pathogens.
Although antibiotic resistance is now a worldwide health threat, there have been few new antibiotics in recent decades. This is where AI offers a solution, both cheaper and less time-consuming than ordinary antibiotic testing.
An AI Antibiotic!
The MIT scientists created an AI system with neural networking capabilities (that mimics neurons in the human brain). This means the AI can "learn" and then "predict" which molecules make the best antibiotics.
The scientists trained the AI with a database of 2,500 known molecules that have antibacterial properties, from which it "learned" what type of molecules are ideal for killing pathogens.
The now-trained AI was given another database of 6,000 molecules and asked to "predict" molecules that are different from the antibiotics we know today.
The system identified a new antibiotic that was then tested on mice and found to be successful in wiping out pathogens such as Enterobacteriaceae and Acinetobacter baumannii — both of which are marked critical-to-target by the World Health Organization. The new antibiotic has been named halicin.
Following their success with halicin, the MIT team fed the AI with 100 million digital molecules from another database, and the AI took three days to find an additional 23 potential antibiotics. Researchers are now testing two of the possible treatments.
Meanwhile, the MIT team plans to improve the AI for the future, hoping to have it discover antibiotics that would only target bad bacteria while leaving the good bacteria intact. They even hope to have the AI create entirely new antibiotics on its own!
Check out this video below about how neural networks learn and predict...
Sources: Nature, The Guardian, Genetic Engineering & Biotechnology News