MULTI-LABEL DISEASE PREDICTION BASED ON GUT MICROBIOME

Abstract: In this study, we will be exploring the gut microbiome of Latin American immigrants to determine what factors of their gut microbiome affect metabolic diseases. The goal of our project is to determine what metabolic diseases/disorders an individual has based on their gut microbiome and other supporting information on the individual. To achieve our goal, we will be exploring machine learning and data analysis techniques to summarize the key points of the data and understand the patterns and relationships in the data.

Retrieving the data locally:

(1) Download the data files from the following Google Drive: https://drive.google.com/drive/folders/1cpUvpXbh3YEHHaW4jmeKhL8DYfE7tE5V?usp=sharing

(2) Place files in data/raw directory

Activating Qiime2

After launching container, open terminal and type in the following command before running run.py:

conda activate qiime2-2022.11

To use within jupyter notebook also type in the following commands:

pip install -–user ipykernel

python -m ipykernel install -–user -–name=qiime2-2022.11

then refresh jupyter hub

and select the qiime2 kernel

Running the Project:

• To revert to a clean repository, from the project root dir, run python run.py clean
  • This deletes all built files
• To run the entire project on test data, from the project root dir, run python run.py test
  • This fetches the test data, creates features, cleans the data, performs permanova tests, creates pcoa plots, creates machine learning model and model performance graphs for given disease types
• To run the entire project on the real data, from the project root dir, run python run.py all
  • This fetches the original data, creates features, cleans the data, performs permanova tests, creates pcoa plots, performs UMAP, creates machine learning model and model performance graphs for given disease types

Model Performance

To view model performance graphs, permanova tests, and pcoa plots after running run.py, download .qzv files from data/out and upload to https://view.qiime2.org/

Collaborator: Amando Jimenez, Emerson Chao, Renaldy Herlim

For more information visit: https://renaldyh27.github.io/Capstone-Website/