P36 Using machine learning to identify and stratify patients with juvenile-onset SLE

Background Juvenile-onset Systemic Lupus Erythematosus (JSLE) is a complex disease characterised by diagnosis and treatment delays. We applied a machine learning (ML) approach to explore new diagnostic signatures for JSLE based on immune-phenotyping data. Methods Immune-phenotyping of 28 T-cell, B-cell and myeloid-cell subsets in 67 age and sex-matched JSLE patients and 39 healthy controls (HCs) was performed by flow cytometry. A balanced random forest ML predictive model was developed (10,000 decision trees). 75% of sample data was randomly selected as a training set, the remaining 25% out-of-bag data was used for validation. Reciever operator characteristic, 10-fold cross validation, Sparse Partial Least Squares-Discriminant Analysis (sPLS-DA) and linear regression was used to validate the model. Results In JSLE, a global change in immunological architecture was established compared to HCs: many of the immune cell relationships identified in HCs using correlation comparison analysis were inverted or exacerbated in JSLE, including significantly inverted correlations between intermediate monocytes and memory B-cell populations and CD4/CD8 memory T-cells and B-cell memory in SLE versus HCs. Using immune-phenotyping data a ML model was developed and validated (accuracy=87.80%) showing that JSLE patients could be distinguished from HCs with high confidence using immunological parameters. The top variables contributing to the model included CD19+ unswitched memory B-cells, naive B-cells, CD14+ monocytes and memory T-cell subsets. The ‘JSLE immune signature’ was also sucessfully validated using sPLS-DA and linear regression. To assess whether the validated signature could be used to further stratify JSLE patients, K-mean clustering was applied. Four JSLE groups each with a distinct immune and clinical profile were identified. Finally, network analysis identified specific clinical features associated with each of the top JSLE immune-signature variables. Conclusion Using a combined ML approach, a distinct immune signature was identified that discriminated between JSLE patients and HCs and further stratified patients. This signature could have diagnostic and therapeutic implications.