Trans-omics analysis of post-injury thromboinflammation plasma identifies endotypes and trajectories in trauma patients
Data files
Aug 11, 2025 version files 13.26 MB
-
Data.file.S1.data_COMBAT.xlsx
12.77 MB
-
Multiomics.analysis.Rmd
481.90 KB
-
README.md
4.44 KB
Abstract
Understanding and managing the complexity of trauma-induced thromboinflammation necessitates an innovative, data-driven approach. This study leveraged a trans-omics analysis of 759 longitudinal samples from 118 trauma patients and 97 healthy controls to illuminate molecular endotypes and trajectories that underpin patient outcomes. We hypothesized that unsupervised trans-omics profiling would reveal underlying clinical differences in injured patients that may present with similar clinical characteristics but ultimately have different outcomes. Here, we used proteomics and metabolomics to profile longitudinal plasma samples from trauma patients and healthy controls. Omics-based patient states were defined to map unique pathophysiologic states encountered by trauma patients over time. Then, patients were endotyped according to their longitudinal trajectory through trauma omics states, and injury patterns and outcomes were compared. Importantly, endotypes without significant differences in injury patterns yet with different clinical outcomes were identified. Organ failure among these similarly injured patients was predicted with higher accuracy using omics markers over injury covariates. Patients who presented with elevated proteosome activation, catabolism, and superoxide formation were vulnerable to heart and lung failure, and ALI, respectively. Additionally, hypoxia, RBC lysis, and hydrolase omics markers out-predicted injury covariates for mortality and intensive care across all trauma patients. Injury and outcome patterns persisted in an independent validation cohort of 333 patients from the Trauma Activation Protocol trial following trajectory prediction using a single, early timepoint. This strategy aligns with our understanding that trauma patients, despite similar clinical presentation, might harbor vastly different biological responses and outcomes. Further, this work presents a novel framework for personalized trauma patient treatment by mapping patient trajectory through injury and recovery.
https://doi.org/10.5061/dryad.d51c5b0dn
Description of the data and file structure
Plasma was collected from trauma (severely injured) patients, which was processed for and quantified by mass spectrometry-based proteomics and metabolomics. Data underwent supervised analyses to compare differences among hemorrhagic shock and tissue injury groups, followed by unsupervised analyses to determine omics-driven endotypes to determine differences in trauma severity and clinical outcomes among these sub-phenotypes. In this directory is Data.file.S1.data_COMBAT.xlsx which contains proteomics and metabolomics data (clinical data available upon request) as well as .Rmd file which contains R scripts that were used for modeling and analyses.
Files and variables
File: Table.S16.data_COMBAT.xlsx
Description: Proteomics and Metabolomics data from trauma patient plasma. Rows are individual patient*timepoints, and column names are gene names for proteins quantified, as well as metabolites quantified by mass spectrometry. HC indicated a healthy control. C## indicates patient number (which is anonymized). Hour### indicates the time point at which the plasma was collected
Variables
- Each column name is the gene name for the protein quantified (ie, IGKV3_7 is Immunoglobulin Kappa Variable 3-7). Column names for metabolomics are metabolite names.
- Values at or below the limit of detection are 0
This project performs unsupervised cluster analysis on plasma collected from trauma patients longitudinally. Plasma was subjected to mass spectrometry based proteomics and metabolomics to profile and quantify identified markers. Raw proteomics MS files are available on PRIDE (PXD053799), and raw metabolomics MS files are available on Metabolomics Workbench ([https://dx.doi.org/10.21228/M85R82]. Data are available here after searching on Spectronaut using a project-specific DDA library for proteomics and MAVEN for metabolomics, and are presented as relative units of intensity.
NOTE: Data here only include proteomics and metabolomics. Clinical data may be made available upon request.
- data_COMBAT.xlsx contains proteomics and metabolomics data, and
- Multiomics.analysis.Rmd contains R scripts for processing and analyzing data used for this manuscript.
R software (4.4.0) with Tidyverse (2.0.0) and the following packages were used for data analysis and generating graphs. broom (1.0.6), reshape2 (1.4.4), readxl (1.4.3), openxlsx (4.2.7.1), and xlsx2dfs (0.1.0) were also used for data management. UMAP (uwot 0.2.2) was used to create an embedding of the omics manifolds followed by hierarchical clustering to identify patient states. Nbclust (3.0.1) and clValid (0.7) were used to determine cluster number. Pheatmap (1.0.12) generated heatmaps. Metaboanalyst and Metascape (using KEGG, PANTHER, STRING, MCODE, and DisGeNET databases) websites were used for pathway enrichment. ggbreak (0.1.2), ggpubr (0.6.0), plotly (4.10.4), gplots (3.1.3.1), ggplot2 (3.5.1), RColorBrewer (1.1.3), scales (1.3.0), svglite (2.1.3), viridis (0.6.5), gridExtra (2.3), plotrix (3.8-4), and ggsignif (0.6.4) were used for graphing. mixOmics (6.28.0), randomForest (4.7-1.1), caret (6.0-94), glmnet (4.1-8), rstatix (0.7.2), pROC (1.18.5), xgboost (1.7.8.1), biglasso (1.6.0), FactoMineR (2.11), gtools (3.9.5), MLeval (0.3), ROCR (1.0-11), dunn.test (1.3.6), factoextra (1.0.7), logistf(1.26.0), EnhancedVolcano (1.22.0), ranger (0.16.0), KernelKnn (1.1.5), arm (1.14-4), MASS (7.3-61), and matrixTests (0.2.3) were used for data analysis and modeling. lme4 (1.1-35.5) and emmeans (1.10.4) were used for linear mixed modeling. recipes (1.1.0) was used for Yeo-Johnson data transformation prior to comparing COMBAT and TAP validation datasets. Dynamic time warping (dtw 1.23-1) was used for trajectory divergence calculation. circlize (0..4.16) was used for creating cord diagrams.
Human subjects data
During enrollment for this study, explicit consent was obtained from each participant for the use and publication of their de-identified data into the public domain. This data is de-identified through the use of an anonymized identifier, and contains no personally identifiable information.
Data are proteomics and metabolomics intensity values from trauma patient plasma processed via mass spectrometry. For proteomics, plasma samples were tryptic digested using S-Trap 96-well plates. Peptides were lyophilized, resuspended in 0.1% formic acid, and loaded onto Evotips. Peptides were analyzed using the Evosep One system coupled to a timsTOF Pro mass spectrometer (diaPASEF mode) via the nano-electrospray ion source. Raw DIA files were searched in Spectronaut using a project-specific spectral library, and data were presented as units of relative intensity.
For metabolomics, frozen plasma aliquots (10 µL) were extracted 1:25 in ice-cold extraction solution (methanol:acetonitrile: water 5:3:2 v/v/v). Samples were vortexed for 30 min at 4℃ prior to centrifugation for 10 min at 15,000g at 4℃. Analyses were performed using a Vanquish UHPLC coupled online to a Q Exactive mass spectrometer (ThermoFisher). Samples were analyzed using a 1-minute and 5-minute gradient-based method, spectra were searched in Maven, and data were presented as units of relative intensity.