Researchers have developed a new method to reverse engineer molecules — groups of atoms that make up nearly every physical material, including medications and batteries — using generative AI, according to a new paper published in Nature.
Historically, many therapies have been discovered by trial-and-error or accident, with the most famous example being penicillin. As computational models improved, scientists increasingly tried reverse-engineering molecules: Rather than starting with different molecular structures and testing what they did, they began by specifying how they want molecules to behave — called “properties” — and asking technology to suggest potential structures to achieve that.
Now, the team led by researchers at New York University and the University of Florida says its neural network proposes potentially viable molecular structures at 10 times the speed of existing methods, without sacrificing accuracy. That lets scientists run experiments and develop real-world applications more quickly.