Revising the diagnosis of idiopathic uveitis by peripheral blood transcriptomics.

Abstract Purpose To test the hypothesis that idiopathic uveitis can be categorized into subtypes based on gene expression from blood. Design Case control study Methods We applied RNA-Seq to peripheral blood from patients with uveitis associated with one of four systemic diseases, including axial spondyloarthritis (n=17), sarcoidosis (n=13), inflammatory bowel disease (n=12), tubulo-interstitial nephritis with uveitis (n=10), or idiopathic uveitis (n=38) as well as 18 healthy controls evaluated predominantly at Oregon Health & Science University. A high dimensional negative binomial regression model implemented in the edgeR R package compared each disease group against the controls. The 20 most distinctive genes for each diagnosis were extracted. Out of 80 genes, there were 75 unique genes. A classification algorithm was developed by fitting a gradient boosting tree with 5-fold cross-validation. mRNA from subjects with idiopathic uveitis was analyzed to see if any fit clinically and by gene expression pattern with one of the diagnosable entities. Results For uveitis associated with a diagnosable systemic disease, gene expression profiling achieved an overall accuracy of 85% (balanced average of sensitivity plus specificity, p-value Conclusions Peripheral blood gene expression profiling is a potential adjunct in accurate differential diagnosis of the cause of uveitis. Validation of these results and characterization of the gene expression profile from additional discrete diagnoses could enhance the value of these observations.