Facebook has claimed that its new artificial intelligence, known as 'Compositional Perturbation Autoencoder' (CPA) is able to predict how drugs interact inside cells faster than traditional lab methods and that its new system will greatly help researchers in finding treatments for conditions like cancer.
While repurposing existing drugs has proven to be useful in treating diseases like cancer, figuring out which drug combos will work is a difficult task. Given the vast number of available drugs out there, without software, it is extremely challenging to figure out which drugs work best together. For example, with a pool of just 100 drugs, researchers would need to try out 500 to 19 billion solutions to find the best drug regimen.
The AI system developed by Facebook AI Research and Helmholz Centre in Munich, Germany works by measuring how individual cells respond after being treated by certain drugs. To produce results, the AI takes into account aspects including the type of drug used and dosage, how it interacts with other drugs, the time it takes to work, the type of cell it targets, and other interventions like gene deletion. Using this information, it is then able to predict how different drug combinations it is not familiar with may affect cells.
Unlike supervised AI models that learn from labeled datasets, this new model generates labels from data on the relationships between the data’s parts. And unlike in traditional trials and experiments, predictions take just hours to generate as opposed to years. In tests, the researchers found that the AI was able to forecast cell responses with more than 90% accuracy in known drug combinations. Of course, though, the more the drugs were familiar to the AI beforehand, the more accurate the results were.
“Our hope is that pharmaceutical and academic researchers as well as biologists will utilize [CPA] to accelerate the process of identifying optimal combinations of drugs for various diseases,” wrote Facebook program manager, Anna Klimovskaia, and research scientist, David Lopez-Paz, in a blog post.
“In the future, [CPA] could not only speed up drug repurposing research, but also - one day- make treatments much more personalized and tailored to individual cell responses, one of the most active challenges in the future of medicine to date.”