Search

There is considerable biological and physiological heterogeneity among patients who meet standard clinical criteria for acute respiratory distress syndrome (ARDS). In this study, we tested the hypothesis that there exists a subgroup of ARDS patients who exhibit a metabolically distinct profile. We examined undiluted pulmonary edema fluid obtained at the time of endotracheal intubation from 16 clinically phenotyped ARDS patients and 13 control patients with hydrostatic pulmonary edema. Nontargeted metabolic profiling was carried out on the undiluted edema fluid. Univariate and multivariate statistical analyses including principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were conducted to find discriminant metabolites. Seven-hundred and sixty unique metabolites were identified in the pulmonary edema fluid of these 29 patients. We found that a subset of ARDS patients (6/16, 38%) presented a distinct metabolic profile with the overrepresentation of 235 metabolites compared with edema fluid from the other 10 ARDS patients, whose edema fluid metabolic profile was indistinguishable from those of the 13 control patients with hydrostatic edema. This “high metabolite” endotype was characterized by higher concentrations of metabolites belonging to all of the main metabolic classes including lipids, amino acids, and carbohydrates. This distinct group with high metabolite levels in the edema fluid was also associated with a higher mortality rate. Thus metabolic profiling of the edema fluid of ARDS patients supports the hypothesis that there is considerable biological heterogeneity among ARDS patients who meet standard clinical and physiological criteria for ARDS. [ABSTRACT FROM AUTHOR]

Background: Metabolic predictors and potential mediators of survival in sepsis have been incompletely characterized. We examined whether machine learning (ML) tools applied to the human plasma metabolome could consistently identify and prioritize metabolites implicated in sepsis survivorship, and whether these methods improved upon conventional statistical approaches. Methods: Plasma gas chromatography–liquid chromatography mass spectrometry quantified 411 metabolites measured ≤ 72 h of ICU admission in 60 patients with sepsis at a single center (Brigham and Women's Hospital, Boston, USA). Seven ML approaches were trained to differentiate survivors from non-survivors. Model performance predicting 28 day mortality was assessed through internal cross-validation, and innate top-feature (metabolite) selection and rankings were compared across the 7 ML approaches and with conventional statistical methods (logistic regression). Metabolites were consensus ranked by a summary, ensemble ML ranking procedure weighing their contribution to mortality risk prediction across multiple ML models. Results: Median (IQR) patient age was 58 (47, 62) years, 45% were women, and median (IQR) SOFA score was 9 (6, 12). Mortality at 28 days was 42%. The models' specificity ranged from 0.619 to 0.821. Partial least squares regression-discriminant analysis and nearest shrunken centroids prioritized the greatest number of metabolites identified by at least one other method. Penalized logistic regression demonstrated top-feature results that were consistent with many ML methods. Across the plasma metabolome, the 13 metabolites with the strongest linkage to mortality defined through an ensemble ML importance score included lactate, bilirubin, kynurenine, glycochenodeoxycholate, phenylalanine, and others. Four of these top 13 metabolites (3-hydroxyisobutyrate, indoleacetate, fucose, and glycolithocholate sulfate) have not been previously associated with sepsis survival. Many of the prioritized metabolites are constituents of the tryptophan, pyruvate, phenylalanine, pentose phosphate, and bile acid pathways. Conclusions: We identified metabolites linked with sepsis survival, some confirming prior observations, and others representing new associations. The application of ensemble ML feature-ranking tools to metabolomic data may represent a promising statistical platform to support biologic target discovery. [ABSTRACT FROM AUTHOR]

Background: Metabolic predictors and potential mediators of survival in sepsis have been incompletely characterized. We examined whether machine learning (ML) tools applied to the human plasma metabolome could consistently identify and prioritize metabolites implicated in sepsis survivorship, and whether these methods improved upon conventional statistical approaches. Methods: Plasma gas chromatography–liquid chromatography mass spectrometry quantified 411 metabolites measured ≤ 72 h of ICU admission in 60 patients with sepsis at a single center (Brigham and Women's Hospital, Boston, USA). Seven ML approaches were trained to differentiate survivors from non-survivors. Model performance predicting 28 day mortality was assessed through internal cross-validation, and innate top-feature (metabolite) selection and rankings were compared across the 7 ML approaches and with conventional statistical methods (logistic regression). Metabolites were consensus ranked by a summary, ensemble ML ranking procedure weighing their contribution to mortality risk prediction across multiple ML models. Results: Median (IQR) patient age was 58 (47, 62) years, 45% were women, and median (IQR) SOFA score was 9 (6, 12). Mortality at 28 days was 42%. The models' specificity ranged from 0.619 to 0.821. Partial least squares regression-discriminant analysis and nearest shrunken centroids prioritized the greatest number of metabolites identified by at least one other method. Penalized logistic regression demonstrated top-feature results that were consistent with many ML methods. Across the plasma metabolome, the 13 metabolites with the strongest linkage to mortality defined through an ensemble ML importance score included lactate, bilirubin, kynurenine, glycochenodeoxycholate, phenylalanine, and others. Four of these top 13 metabolites (3-hydroxyisobutyrate, indoleacetate, fucose, and glycolithocholate sulfate) have not been previously associated with sepsis survival. Many of the prioritized metabolites are constituents of the tryptophan, pyruvate, phenylalanine, pentose phosphate, and bile acid pathways. Conclusions: We identified metabolites linked with sepsis survival, some confirming prior observations, and others representing new associations. The application of ensemble ML feature-ranking tools to metabolomic data may represent a promising statistical platform to support biologic target discovery. [ABSTRACT FROM AUTHOR]

Using latent class analysis (LCA) of clinical and protein biomarkers, researchers have identified two phenotypes of the acute respiratory distress syndrome (ARDS) with divergent clinical trajectories and treatment responses. We investigated whether plasma metabolites differed among patients with LCA-derived hyperinflammatory and hypoinflammatory ARDS, and we tested the prognostic utility of adding metabolic clusters to LCA phenotypes. We analyzed data from 93 patients with ARDS and sepsis enrolled in a multicenter prospective cohort of critically ill patients, comparing 970 metabolites between the two LCA-derived phenotypes. In all, 188 metabolites were differentially abundant between the two LCA-derived phenotypes. After adjusting for age, sex, confounding medications, and comorbid liver and kidney disease, 82 metabolites remained significantly different. Patients with hyperinflammatory ARDS had reduced circulating lipids but high levels of pyruvate, lactate, and malate. Metabolic cluster and LCA-derived phenotypes were each significantly and independently associated with survival. Patients with hyperinflammatory ARDS may be experiencing a glycolytic shift leading to dysregulated lipid metabolism. Metabolic profiling offers prognostic information beyond what is captured by LCA phenotypes alone. Deeper biological profiling may identify key differences in pathogenesis among patients with ARDS and may lead to novel targeted therapies. [ABSTRACT FROM AUTHOR]