Interpretable Machine Learning Decodes Soil Microbiome’s Response to Drought Stress

Link to Publication: https://doi.org/10.1186/s40793-024-00578-1

This repository contains the code used for 16S rRNA-based metagenomics analysis and ML applications for drought stress identification in the soil metagenome using a dataset from Naylor et al. (2017) further refered to as the grass-drought dataset.

• SoilMicrobiomeDroughtML/
  • Metagenomics_Analysis.md: Markdown showing how Data Processing, Diversity Analysis and Differential Abundance Analysis was performed on the grass-drought dataset
  • Metagenomics_Analysis_files/: Folder containing plots to be displayed in Metagenomics_Analysis.md
  • Marker_Taxa.md: Markdown showing the comparison of significant taxa from DAA and important taxa from interpretable ML for the grass-drought dataset
  • Marker_Taxa_files/: Folder containing plots to be displayed in Marker_Taxa.md
  • Machine_Learning.ipynb: Nested CV of Random Forest Classifier and interpretable ML with SHAP values with the grass-drought dataset
  • data/
    • metadata.csv: Table containing enrichment information per sample for the grass-drought dataset
    • DADA2_ASV_count.Rdata: Table containing ASV counts per sample of the grass-drought dataset
    • DADA2_ASV_taxonomy.Rdata: Table containing taxonomic annotation per ASV of the grass-drought dataset
    • feature_tbl_{phylum/class/order/family/genus}.csv: Table containing relative abundances of phyla/classes/orders/families/genera per sample with watering regime as target of the grass-drought dataset
  • supplementary_analysis:
    • Logistic_Regression.py: Python script running Nested CV with Logistic Regression Classifier on the grass-drought dataset
    • Machine_Learning_Hold_Out.py: Python script running Nested CV of Random Forest Classifier on the grass-drought dataset after the creation of a hold-out dataset from the grass-drought dataset and predicting on the hold-out dataset

The scripts need to be executed in the following order:

1. Metagenomics_Analysis.Rmd
2. Machine_Learning.ipynb
3. Marker_Taxa.Rmd

Grass-Drought dataset:

Naylor, D., DeGraaf, S., Purdom, E., Coleman-Derr, D.: Drought and host selection influence bacterial community dynamics in the grass root microbiome. The ISME Journal 11(12), 2691--2704 (2017) https://doi.org/10.1038/ismej.2017.118