In Frontiers in medicine
Sarcoidosis is a granulomatous disease of unknown etiology, immunologically characterized by a Th1 immune response. Transcriptome-wide expression studies in various types of sarcoid tissues contributed to better understanding of disease mechanisms. We performed a systematic database search on Gene Expression Omnibus (GEO) and utilized transcriptomic data from blood and sarcoidosis-affected tissues in a meta-analysis to identify a cross-tissue, cross-platform signature. Datasets were further separated into training and testing sets for development of a diagnostic classifier for sarcoidosis. A total of 690 differentially expressed genes were identified in the analysis among various tissues. 29 of the genes were robustly associated with sarcoidosis in the meta-analysis both in blood and in lung-associated tissues. Top genes included LINC01278 (P = 3.11 × 10-13), GBP5 (P = 5.56 × 10-07), and PSMB9 (P = 1.11 × 10-06). Pathway enrichment analysis revealed activated IFN-γ, IL-1, and IL-18, autophagy, and viral infection response. IL-17 was observed to be enriched in peripheral blood specific signature genes. A 16-gene classifier achieved excellent performance in the independent validation data (AUC 0.711-0.964). This study provides a cross-tissue meta-analysis for expression profiles of sarcoidosis and identifies a diagnostic classifier that potentially can complement more invasive procedures.
Jiang Yale, Jiang Dingyuan, Costabel Ulrich, Dai Huaping, Wang Chen
IL-17, interferon, machine learning, sarcoidosis, transcriptome