Study Finds Four Predictive Lupus Disease Profiles Using Machine Learning

A new study using machine learning (ML) identified four distinct lupus disease profiles or autoantibody clusters that are predictive of long-term disease, treatment requirements, organ involvement and risk of death. Machine learning refers to the process by which a machine or computer can imitate human behavior to learn and optimize complicated tasks such as statistical analysis and predictive modeling using large datasets. Autoantibodies are antibodies produced by the immune system and directed against proteins in the body. Proteins are often a cause or marker for many autoimmune diseases, including lupus.

Researchers observed 805 people with lupus, looking at demographic, clinical, and laboratory data within 15-months of their diagnosis, then again at 3-years, and 5-years with the disease. After analyzing the data, the researchers used predictive ML which revealed four distinct clusters or lupus disease profiles associated with important lupus outcomes:

• Cluster 1 (137 people):
  • High risk - characterized by autoantibody reactivities, high disease activity, and immunosuppressant/biologic use were found more commonly among non-white races/ethnicities known to have more severe lupus
• Cluster 2 (376 people):
  • The lowest frequency of lupus nephritis (LN) and lowest immunosuppressant/biologic use
• Cluster 3 (80 people):
  • Highest frequency of antiphospholipid antibodies and predictive of cardiovascular disease events including strokes and antiphospholipid syndrome
• Cluster 4 (212 people):
  • High disease activity characterized by multiple autoantibody reactivity

Further studies are needed to determine other lupus biomarkers and understand disease pathogenesis through ML approaches. The researchers suggest ML studies can also help to inform diagnosis and treatment strategies for people with lupus. Learn more about lupus research.

Read the study