Lupus & antiphospholipid syndrome
Identifying Systemic Lupus Erythematosus From Serum Proteomic Profiles Using Machine Learning and Genetic Risk Stratification
Authors' conclusion
Proteomic machine learning models show excellent performance for identifying pre-existing lupus and generalize well for predicting lupus before clinical diagnosis.
Study context
- Population
- UK Biobank participants with proteomic data, incl. patients with SLE and individuals with other autoimmune diseases; replication cohorts from Sweden (Human Protein Atlas (HPA)) and China383 SLE patients
- Participants
- 44,173 participants
- Comparison
- Proteomic machine learning model vs linear model vs polygenic risk scores
- Outcome
- Classification of established SLE and prediction of future SLE from serum proteomic profiles
Methodological detailsmachine learning, proteomic, replication in independent cohorts (Sweden HPA, China)
- Machine learning
- Proteomic
- Replication in independent cohorts (Sweden HPA, China)
Performance
As reported in the abstract
Identification of lupus (pre-existing and future) from serum proteomic profiles
Sensitivity at 95% specificity among patients with lupus on immunomodulatory medications (replicated in an independent cohort)
≈90%
Further findings
Replication of SLE proteomic associations in HPA
- Lupus proteomic associations replicated, with moderate correlation in effect sizes
- ≈70%
Dysregulated proteins
- SLE showed the largest number of dysregulated proteins among autoimmune diseases and clustered with RA
Model interpretation
- SCARB2, SOD2, CD302, Galectin-9 and GGT5 had substantial effects on SLE identification
Original publication
Identifying Systemic Lupus Erythematosus From Serum Proteomic Profiles Using Machine Learning and Genetic Risk Stratification. Arthritis Rheumatol. Published 4 May 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →