A machine learning approach for precision diagnosis of juvenile-onset SLE

Juvenile-Onset systemic lupus erythematosus (JSLE) is an autoimmune rheumatic disease characterised by systemic inflammation and organ damage, with disease onset often coinciding with puberty. JSLE is associated with more severe disease manifestations and a higher motility rate compared to adult SLE. Due to the heterogeneous clinical and immunological manifestations of JSLE, delayed diagnosis and poor treatment efficacy are major barriers for improving patient outcome. In order to define a unique immunophenotyping profile distinguishing JSLE patients from age matched healthy controls, immune-based machine learning (ML) approaches were applied. Balanced random forest analysis discriminated JSLE patients from healthy controls with an overall 91% prediction accuracy. The top-ranked immunological features were selected from the optimal ML model and were validated by partial least squares discriminant analysis and logistic regression analysis. Patients could be clustered into four distinct groups based on the top hits from the ML model, providing an opportunity for tailored therapy. Moreover, complex correlations between the JSLE immune profile and clinical features of disease were identified. Further immunological association studies are essential for developing data-driven personalised medicine approaches to aid diagnosis of JSLE for targeted therapy and improved patient outcomes.