COVID infection has been linked to higher risk of autoimmune disorders, including rheumatoid arthritis and type 1 diabetes. But why the virus might cause the body’s immune system to go haywire remains unknown, making it difficult to develop therapies to avoid autoimmunity. One hypothesis is that viral “molecular mimics” that resemble the body’s own proteins trigger an immune response against the virus — and healthy tissues get caught in the crossfire.
Now, with advanced data analysis and machine learning, scientists have identified a set of COVID-derived molecular mimics that are most likely to be involved in triggering autoimmunity.
The new results are published in ImmunoInformatics.
The researchers first looked for viral components that are similar to the human proteins known to be attacked in various autoimmune diseases. Theoretically, these viral proteins could trigger the immune system to target the human proteins they resemble. They narrowed down their list of culprits by using machine learning to identify only those viral components that are most likely to be bound by human antibodies.
Some of the viral components the researchers found have been associated with type 1 diabetes or multiple sclerosis.
Importantly, some of the human proteins that the researchers identified as likely targets of COVID-induced autoimmunity are only found in people with specific genetics, suggesting that people who produce those proteins may be at higher risk of COVID-induced autoimmunity.
“It’s exciting that in collaboration with our clinical colleagues, we can now use AI and machine learning to address medical conditions exacerbated by the COVID pandemic,” says Julio Facelli, PhD, distinguished professor of biomedical informatics at University of Utah Health and the senior author on the paper. “Hopefully, our results will lead to better understanding and eventual treatment and prevention of these debilitating conditions.”
The results are published in Immunoinformatics as “Molecular mimicry impact of the COVID-19 pandemic: Sequence homology between SARS-CoV-2 and autoimmune diseases epitopes.”
Research was supported by the National Library of Medicine (5T15LM007124-24) and by the CTSA award to the Utah Clinical and Translational Science Institute (UM1TR004409). Computational resources were provided by the Utah Center for High Performance Computing. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.