This repository contains the following directories.

Genome Denoising

TODO

Phenotype Prediction

This directory contains implementations of the phenotype prediction and feature selection pipelines. The directory structure is the following:

• data_preparation: contains pipeline and data for test/train data split,
• data_[phenotype]: input/output data directory for [phenotype],
• jupyter_notebooks: contains notebooks with the scripts for phenotype prediction and feature selection for each phenotype.

At the moment, there are three phenotypes:

• aerobicity,
• OGT,
• mono/didermy.

First, the input test/train datasets should be generated for the chosen phenotype. The split is done at a chosen taxonomy level (i.e. samples from the same taxonomy group are not split between train and test). To generate the input splits, run the following (see taxa_level_split.py description for more details)

cd ~/gene-context/phenotype_prediction
python3 -m data_preparation.taxa_level_split \ --tax_level [tax_level] \ --input_annotation_csv [input_annotation_csv] \ --input_data_csv [input_data_csv] \ --output_dir [output_dir]

Please, note that ar122_metadata_r202.tsv and bac120_metadata_r202.tsv are required to be stored in gene-context/phenotype_prediction/data_preparation/gtdb_files in order to run the above command.

After the inputs are generated, a notebook for the corresponding [phenotype] can be run.