Pluto Bioinformatics

GSE133298: Longitudinal Blood Transcriptomic Changes Predict Lung Function Decline in patients with Idiopathic Pulmonary Fibrosis

Bulk RNA sequencing

Rationale: Molecular markers of disease progression in idiopathic pulmonary fibrosis are needed.; Objective: Derive and validate a blood transcriptomic predictor of forced vital capacity (FVC) decline.; Methods: A training cohort (n=74) of IPF patients was stratified according to the presence of progressive disease, defined as 10% relative decline in FVC over 12 months. Baseline to 4-month within-patient changes in gene expression were correlated with categorical FVC decline. Genes predictive of FVC decline were identified by two-group comparison with false discovery rate <5% followed by logistic LASSO regression and 10-Fold Cross-Validation for gene list prioritization. Independent validation cohorts with differing transcriptome assay platforms and blood transcriptome sampling times from UChicago (n=27), UPMC (n=35), and Imperial (n=24) underwent receiver operating characteristic with area under the curve (AUC) analyses for validation.; Results: A longitudinally-derived FVC-gene predictor accurately discriminated most patients with stable and progressive IPF across four independent IPF cohorts with variable transcriptomic assay platforms and sampling times. The FVC-gene predictorand demonstrated sensitivity and specificity of 74.3% and 82.4% in the combined replication cohort. The likelihood ratio, LR+ and LR- were 4.11 and 0.32, respectively. TGF-beta was the highest-ranking canonical pathway by Gene Set Enrichment Analysis. An approach using longitudinal gene expression changes approach dramatically reduced within-group variation compared to cross-sectional expression for improved prediction modeling.; Conclusions: This novel FVC-gene predictor developed from short-term longitudinal gene expression changes successfully discriminates most patients with high likelihood of one-year 10% FVC decline. This tool may better reflect disease activity and prove useful for predictive enrichment of clinical trial populations. SOURCE: Yong Huang (yh9fj@virginia.edu) - University of Viginia