AI examines gene-expression data to predict COVID-19 outcomes

Editor's Note

In this study, researchers from the University of California, San Diego, developed an artificial intelligence (AI) algorithm to identify signature genes that forecast the severity of a viral infection immune response including that of COVID-19.

The researchers analyzed more than 45,000 datasets from viral pandemics in humans, mice, and rats. They identified a 166-gene signature that predicts how the immune system reacts to viral infections and a signature of 20 genes that can predict the severity of a patient’s condition.

The 166-gene signature revealed a phenomenon termed “cytokine storm,” in which the body releases too many cytokines that causes the immune system to attack healthy tissue. The researchers were able to define the source of the cytokine storm and found that the storm can lead to lung airway cell damage, which prevents the immune system from destroying cells infected with the virus.

The authors say that as new COVID-19 datasets emerge, the AI model will be even more effective and accurate. The gene signatures also performed well with identifying and classifying bacterial and fungal infections, which shows promise for its effectiveness outside of COVID-19.