Blood transcriptomics to facilitate diagnosis and stratification in pediatric rheumatic diseases – a proof of concept study

Abstract

Background

Transcriptome profiling of blood cells is an efficient tool to study the gene expression signatures of rheumatic diseases. This study aims to improve the early diagnosis of pediatric rheumatic diseases by investigating patients’ blood gene expression and applying machine learning on the transcriptome data to develop predictive models.

Methods

RNA sequencing was performed on whole blood collected from children with rheumatic diseases. Random Forest classification models were developed based on the transcriptome data of 48 rheumatic patients, 46 children with viral infection, and 35 controls to classify different disease groups. The performance of these classifiers was evaluated by leave-one-out cross-validation. Analyses of differentially expressed genes (DEG), gene ontology (GO), and interferon-stimulated gene (ISG) score were also conducted.

Results

Our first classifier could differentiate pediatric rheumatic patients from controls and infection cases with high area-under-the-curve (AUC) values (AUC = 0.8 ± 0.1 and 0.7 ± 0.1, respectively). Three other classifiers could distinguish chronic recurrent multifocal osteomyelitis (CRMO), juvenile idiopathic arthritis (JIA), and interferonopathies (IFN) from control and infection cases with AUC ≥ 0.8. DEG and GO analyses reveal that the pathophysiology of CRMO, IFN, and JIA involves innate immune responses including myeloid leukocyte and granulocyte activation, neutrophil activation and degranulation. IFN is specifically mediated by antibacterial and antifungal defense responses, CRMO by cellular response to cytokine, and JIA by cellular response to chemical stimulus. IFN patients particularly had the highest mean ISG score among all disease groups.

Conclusion

Our data show that blood transcriptomics combined with machine learning is a promising diagnostic tool for pediatric rheumatic diseases and may assist physicians in making data-driven and patient-specific decisions in clinical practice.