Multi-omics Analysis for Diagnosis of Post-trauma Blood Clots

Lazarina Butkovich, Selina Jessica, Vansika K, Nadia Tasevski

Open Bootcamp Collective, Project 2 - 2025

Study: Zhang et al. "Integrated landscape ofplasma metabolism and proteome ofpatients with post-traumatic deep vein thrombosis" Nature Communications (2025). https://doi.org/10.1038/s41467-024-52262-0

Open Bootcamp Collective Presentation: Google Slides

Objectives

• For this portfolio bioinformatics project, we replicate analyses of Zhang et al.
• Metabolomics, single-omics
• Proteomics, single-omics (in progress)
• Clinical Parameters (in progress)
• Multi-omics (in progress)
• Predictive Model for pt-DVT (in progress)

Background

• Study Design
  • Zhang et al. collected plasma metabolomics and proteomics data for 680 individuals, with and without deep vein thrombosis (DVT) (aka blood clot) after trauma, referred to as post-trauma deep vein thrombosis (pt-DVT).
  • DVT is a major health problem that can lead to complication, including post-thrombotic syndrome, recurrent DVT, and life-threatening pulmonary embolism. Risk of DVT is elevated post-trauma (aka physical injury).
  • Failure to diagnose pt-DVT is a leading cause of death post-trauma. Through their multi-omics analysis, Zhang et al. proposed (i) protein and metabolite biomarkers for early diagnosis of pt-DVT and (ii) potential therapeutic strategies for pt-DVT.
  • In order to develop a predictive model for pt-DVT, Zhang et al. recruited patients for a discovery group (N=580, with n_pt_DVT=252 and n_without_DVT=328 ) and a separate validation group (N=100, with n_pt_DVT=50, n_without_DVT=50).
• Metabolomics Dataset
  • After MS/MS identification, 326 metabolites were determined for metabolomic analysis in all 580 samples of the discovery cohort.
  • To process the data:
    • Process analytical batches separately (batch 1 and batch 2). By scaling by batch, the sensitivity differences between batches is corrected, and the relative abundances matrix of all peaks becomes comparable.
    • Filter to remove noise
      • Remove peaks whose RSD (relative standard deviation) is 40%< in QC samples
      • Remove peaks with more than 50% null values in a single group or all groups
    • Impute data:
      • For remaining features with missing intensity values, replace with half the minimum value for that metabolite across all samples.
    • Normalize by dividing intensities by the intensity of the internal standard (IS) (see data column "ID", value "IS" in Supplementary Data 2; we normalized by the positive ionization "IS")
    • Log-transform
    • Scale by median centered (we scaled via auto scaling - mean-center then divide by the standard deviation of each variable)
  • Use Supplementary Data 3 to directly compare metabolomics data processing of Zhang et al. to the results in this portfolio project.

Scripts Overview for Metabolomics Analysis - Lazarina Butkovich

1. format_inputs_for_analyst.py
   • formats both metabolomic and proteomic data from edited version of Zhang et al Supplementary Datasets (Supplementary Datasets 2 and 13, respectively)
   • Input files made available with Google Drive:
     • metabolites_raw.xlsx
     • proteins_raw.xlsx
     • metabolites_metadata.xlsx
2. MetaboAnalyst
   • Option 1: manually run MetaboAnalyst web platform
     • Use MetaboAnalyst web platform to normalize data per batch from script "format_inputs_for_analyst.py"
     • The outputs of these steps are supplied for the script "combine_normalized_batches.py"
   • Option 2: run MetaboAnalystR
     • metaboAnalystR_data_processing.r (in progress)
3. combine_normalized_batches.py
   • Input files made available with Google Drive:
     • data_normalized_batch_1.csv
     • data_normalized_batch_2.csv
   • Run the combined, normalized data again through the MetaboAnalyst web platform, "Statistical Analysis [one factor]" (skip normalization steps)