Facebook- tool detects cancer and covid- 19 in record time22. April 2021
Facebook- tool detects cancer and covid- 19 in record time
New York, 4/22/2012
Facebook today unveiled a new machine learning tool that it says will speed up the discovery of drug combinations that target cancer, COVID-19 and other diseases.
The company has made the model, called Compositional Perturbation Autoencoder (CPA), which it developed with the Helmholtz Center in Germany, open source.
It is the first single AI model that can predict “the effects of drug combinations, dosages, timing, and even other types of interventions, such as gene knockout or deletion,” according to Facebook AI Research.
The tool aims to accelerate the discovery of tailored drug cocktails for diseases that often vary from person to person and can take years to develop. It works by measuring and recording how certain cells change after drug treatment.
Researchers tested the model’s predictions against currently known drug combinations; it was more than 90% accurate in predicting the correct cell responses.
CPA is now available with an application programming interface and Python package for developers. The method is described in a pre-print paper on bioRxiv.