https://github.com/barbarahelena/ckd-metabolomics

https://github.com/barbarahelena/ckd-metabolomics

Science Score: 13.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
    Found codemeta.json file
  • .zenodo.json file
  • DOI references
  • Academic publication links
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (6.9%) to scientific vocabulary
Last synced: 9 months ago · JSON representation

Repository

Basic Info
  • Host: GitHub
  • Owner: barbarahelena
  • Language: R
  • Default Branch: main
  • Size: 1.62 MB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created over 2 years ago · Last pushed over 1 year ago
Metadata Files
Readme

README.md

HELIUS: Metabolome profiles in chronic kidney disease

Author: Barbara Verhaar, b.j.verhaar@amsterdamumc.nl

Summary

File structure

The scripts in this repo include the following:

Descriptives

CKD subjects and controls were selected from the HELIUS cohort based on the following criteria:
- ACR-KDIGO albuminuria stage A1 (controls) or A2 (CKD)
- Availability of plasma, urine and fecal samples

Machine learning analyses

For the machine learning analyses, a XGBoost algorithm was used in a nested cross-validated design with 100 iterations. The script for these analyses can be found in the XGBoost.py file. For each model, the study population was divided in a training (80%) and test (20%) set. For each iteration, the model was trained on the training set on which the hyperparameters were optimized in a 5-fold CV. The resulting model was tested on the test set. The parameter grid can be found in param_grid_medium.json in the scripts/xgboost folder. Each model resulted in a ROC and feature importance of predictors, averaged over 200 iterations.

In the process_model_results_class.R file, the functions from the functions.R file are used to generate feature importance plots, and violin plots for best predictors to assess differences in concentrations between groups.

The following models were used:
- Plasma metabolite profiles non-diabetic CKD versus healthy controls: AUC 0.59 - Urine metabolite profiles non-diabetic CKD versus healthy controls: AUC 0.65

Plasma metabolites

ROC CKD{width=400px}

Violin plots best predictors CKD

Urine metabolites

ROC CKD{width=400px}

Violin plots best predictors sensitivity model

Regression analyses with highest ranked predictors

Overlap plasma and urine metabolites

Owner

  • Name: Barbara Verhaar
  • Login: barbarahelena
  • Kind: user
  • Location: Amsterdam

PhD candidate @ Amsterdam UMC Vascular medicine

GitHub Events

Total
  • Push event: 1
Last Year
  • Push event: 1