October 15, 2021 – Machine learning has come a long way in a quarter of a century since a computer nicknamed Deep Blue shocked the world by defeating chess champion Gary Kasparov. Today, when our smartphones have much more computing power than Deep Blue, scientists have trained their gaze on even bigger opponents, including potentially fatal diseases such as cancer, heart disease and COVID-19.
When supercomputers search for new drug cocktails to treat these conditions, scientists can power machines with multiple data from decades of research to help inform the analysis. But the coronavirus is still too new and mutates too fast for scientists to turn to these common strategies.
Researchers at the Massachusetts Institute of Technology have a new way to deal with the lack of data on the new virus. They train computers to execute algorithms modeled after signaling networks in the human brain. Like the brain, these neural networks can “learn” and adapt to rapidly changing information, creating new connections on the go.
To identify drug combinations that could work against COVID-19, the researchers asked their computer neural network to evaluate two things at once.
One is to look for pairs of drugs that together will be more powerful antivirals than either drug. This concept of two more effective drugs is known as “drug synergy”.
The computer also looks for parts of the disease that the drugs target, such as proteins or genetic mutations associated with the condition. The idea behind these two approaches is that machines can “learn” which drug cocktails can have the greatest antiviral power.
In its study,published in the Proceedings of the National Academy of Sciences, researchers at the Massachusetts Institute of Technology have uncovered two potential drug cocktails they have discovered using this approach. One combines remdezivir, which the FDA has already approved for the treatment of COVID-19, and reserpine, a drug for high blood pressure. The other pairing is remdezivir and an experimental drug called IQ-1S, one of a family of drugs used to treat autoimmune diseases such as rheumatoid arthritis.
These drug cocktails have not yet been shown to be effective against COVID-19 in human studies. But the results of the study could help drug developers determine which combinations would make the most sense to test while looking for new treatments.