shap-metabolomics

Code base for the study: "Interpretable machine learning with tree-based shapley additive explanations: application to metabolomics datasets for binary classification."

Link to Preprint

Requirements

• Anaconda
• Python >= 3.8.8
• Jupyter

1. git repository

Clone local copy of git repository

git clone https://github.com/obifarin/shap-iml-metabolomics

(or use a git GUI client of your choice)

2. Environment setup and access jupyter notebooks

Setup python environment (pls-da-shap.yml) in the terminal.

(or use anaconda GUI.)

3. Notebook contents

Name	Description
01_Sex_MTBLS404.ipynb	Discriminating biological sex via urine metabolomics.
02_HighFatDiet_MTBLS547.ipynb	The impact of a high-fat diet on bile acids in the cecum.
03_Adenocarcinoma_ST000369.ipynb	Detecting Adenocarcinoma via serum metabolomics.
04-pubmed-metabolomics.ipynb	Keyword occurrences by year for partial least squares regression, random forest, and gradient boosting in metabolomics publications on PubMed.

Code base notes>

• The anaconda environment for this work: pls-da-shap.yml
• The important models are saved in the folder saved_models.
• PyChemometrics is the folder for the library used in PLS-DA computation for this work. Some PLS-DA code wouldn't run if you don't have them in the same directory from which you run the code.
• Data folder contains the raw data used in this study, as prepared by Mendez et al in his paper.