Early, non-invasive detection of pulmonary vascular dysfunction in acute respiratory distress syndrome

PROJECT SUMMARY. Acute respiratory distress syndrome (ARDS) is a common cause of respiratory failure associated with substantial in-hospital mortality. ARDS patients with pulmonary vascular dysfunction (PVD) are a particularly high-risk subgroup with even higher mortality. However, attempts at targeting interventions to mitigate PVD are hindered by a lack of non-invasive biomarkers to identify PVD early in the pathogenesis of ARDS. In a preliminary secondary analysis of ARDSNet FACTT, a large ARDS cohort with invasive pulmonary vascular function data using pulmonary artery catheters and paired serial plasma samples, we observed individuals with PVD had 11- fold higher mortality and could be accurately discriminated by key differentially expressed plasma proteins. Thus, we hypothesize that we can develop an accurate and parsimonious classifier model integrating circulating biomarkers with clinical data to non-invasively detect PVD in ARDS at early, intervenable time points. To accomplish this, we will leverage two BioLINCC cohorts: ARDSNet FACTT and PETAL VIOLET. In Aim 1 of this proposal, we will perform high throughput proteomics on clinically well-annotated plasma specimens from a larger sample of ARDS patients enrolled in ARDSNet FACTT with gold standard quantification of pulmonary vascular function using pulmonary artery catheterization. We will use machine learning to derive and validate an optimal classifier model integrating proteomics and clinical data to non-invasively detect PVD in ARDS. In Aim 2, we will perform high throughput proteomics on clinically well-annotated plasma specimens from early critically ill patients at risk for ARDS in the PETAL network VIOLET and test the predictive capacity of our non-invasive model for predicting incident ARDS and detecting PVD early in the pathogenesis of ARDS. In an exploratory Aim 3, we will use network medicine to identify the key protein-protein interactions and molecular pathways that underpin adverse pulmonary vascular function in FACTT and progression to ARDS in the early, at-risk cohort from VIOLET. The proposed work will yield a testable classification model for predictive enrichment of future clinical trials of pulmonary vascular-targeted treatments in ARDS and generate a public proteomic dataset to spur future mechanistic translational studies.