KEY POINTS

  • Biomedical researchers at Ohio State University have employed machine learning to find new uses for old drugs
  • The process could speed up discovery and approval, as the drugs have already been tested by the FDA for safety and efficacy
  • Google's DeepMind AI recently demonstrated that artificial intelligence could quickly and accurately perform protein folding simulations

Medical researchers at Ohio State University have received promising results from a project that uses artificial intelligence to suggest new uses for old drugs. The team, led by Dr. Ping Zhang, has published results that suggest machine learning might allow for the fast distribution of already-approved medication for new treatments.

Dr. Zhang’s team was looking at heart disease for their experiment, setting a machine-learning algorithm to work on a database of insurance claims. It returned with nine drugs unrelated to heart disease that might result in better outcomes.

Three of them are already being tested as potential treatments for heart disease, but Dr. Zhang says the six new drugs themselves aren’t the point of the project. More important is the possibility of a broader application of machine learning.

“My motivation is applying this is to find drugs for diseases without any current treatment,” he told Ohio State News. “This is very flexible, and we can adjust case-by-case. The general model could be applied to any disease if you can define the disease outcome.”

Artificial Intelligence A visitor at Intel's Artificial Intelligence (AI) Day walks past a signboard during the event in Bangalore, India, April 4, 2017. Photo: MANJUNATH KIRAN/AFP/Getty Images

The drugs in question have also already been run through clinical trials to prove safety, cutting down the time and resources needed to get FDA approval for their new use. Zhang is skeptical that artificial intelligence could ever fully replace humans in drug development. A combination, he says, is more likely. 

“This work shows how artificial intelligence can be used to ‘test’ a drug on a patient, and speed up hypothesis generation and potentially speed up a clinical trial,” Zhang said. “But we will never replace the physician - drug decisions will always be made by clinicians.”

Machine learning is already expanding the boundaries of biomedical development, with Google's DeepMind AI achieving breathtaking speed and accuracy in protein-folding simulations. While both projects have a ways to go before they’re actually implemented, they’re promising additions to the process of creating and approving life-saving treatments.