Idiopathic Pulmonary Fibrosis Identification and Prediction of Disease Severity via Combination of Different...

Despite extensive research effort, accurate and timely diagnosis and prognosis of fibrotic interstitial lung disease (ILD) continue to present challenges. To this end, a variety of biomarkers have been explored to diagnose fibrotic ILD type and to predict disease progression, including markers from exhaled breath and from blood circulation. Recently, we evaluated the diagnostic and prognostic value of short-lived volatile organic compounds (VOCs) in exhaled breath, markers from routine bloodwork, and metabolites captured longer-term in hair. This study, for the first time, ranks these different types of biomarkers relative to each other via machine learning, with the goal to gauge how markers spanning a wide range of timescales might reflect the disease type and severity, and potentially reveal the underlying physiological processes. A Random Forest was trained with forward feature selection and tested on validation subsets. The results reveal novel combinations of biomarkers at different timescales reflecting pulmonary inflammation, oxidative stress, and metabolic dysregulation, and altogether capable of accurately distinguishing between idiopathic pulmonary function (IPF) vs. connective tissue disease ILD (CTD-ILD) and predicting disease severity as seen with pulmonary function testing (PFT).Clinical Relevance- This proof-of-concept study shows that fibrotic ILD type and disease severity can be predicted via novel combinations of biomarkers spanning a broad range of timescales, offering the possibility to gain insight into the physiological processes underlying the disease.