Peng, Junjie; (2025) Using multi-omic signatures and artificial intelligence (AI)/machine learning (ML) to predict long-term prognosis of patients with juvenile systemic lupus erythematosus. Doctoral thesis (Ph.D), UCL (University College London).

Text Thesis (Junjie Peng).pdf - Accepted Version Access restricted to UCL open access staff until 1 April 2026. Download (9MB)

Abstract

Background Juvenile-onset systemic lupus erythematosus (JSLE, onset<18-years) is a rare rheumatic condition with significantly elevated cardiovascular disease (CVD)-risk driven by chronic inflammatory autoimmune processes. Current methods for assessing CVD-risk have limited utility in JSLE clinical practice due to its complex and heterogenous disease course and low prevalence of traditional CVD-risk factors in younger populations. This study aimed to address the unmet need for better predictive biomarkers to improve early detection of high CVD-risk in JSLE, guide treatment selection, and advance understanding of the JSLE immune/molecular landscape with potential relevance for personalised medicine strategies. Methods Unsupervised clustering analysis was performed to stratify JSLE patients (using the APPLE trial cohort: atorvastatin vs. placebo for atherosclerosis progression in JSLE) by their baseline carotid intima-media thickness (CIMT) and CIMT progression patterns over 36-months. Serum lipid metabolomic and novel IgG autoantibody profiling were used to identify signatures predictive of CIMT-progression (placebo and statin). The immune landscape of JSLE was investigated using 30-colour spectral flow cytometry. Machine learning methods were used for biomarker selection and development of predictive stratification models. Results JSLE patients were stratified into three distinct groups based on baseline CIMT measurements irrespective of treatment allocation. Analysis of CIMT-progression over 36-months identified two CIMT-progression rates (high/low) in the placebo group, while patients treated with atorvastatin had three CIMT trajectories (high/intermediate/low progression), unrelated to patient or disease determinants. A robust metabolomic signature predictive of high CIMT-progression was identified in the placebo arm (AUC=80.7%). Autoantibody profiling of a subcohort of the APPLE trial also identified novel autoantibody signatures predicting CIMT-progression (AUC=87%) and statin responses (AUC=96%) in JSLE. Analysis of the JSLE immune landscape revealed distinct immune-cell subclusters associated with disease activity. 4 Conclusions This study demonstrated the potential of multi-omics data and machine learning approaches to define JSLE heterogeneity and identified novel diagnostic/prognostic biomarkers for CVD-risk and statin treatment response.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item