tbd-project-25

Placeholder repository for Team 25 – update with final project title and description when available

Abstract

A concise summary of the project's goals, the problem it addresses, and its intended audience. This section can include potential use cases and key features.

Download datasets

Download data from https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1009337#sec028 S1 File.

Installation

Python environment

python -m venv metro
pip install torch pandas tqdm pyyaml
pip install -I -e .
...

R environment

mamba install -c conda-forge r-base r-rcpp r-igraph r-biocmanager

BiocManager::install("mixOmics")

Quick Start

Given a metabolic or transcriptomic profile, MeTrO can encode either to a mutually shared latent space and then decode it into the other (or back to the original in a probabilistic manner).

The following workflow is planned:

import my_project

model = my_project.load_model(state_dict.pkl)
z = model.encode(metabol_profile)
recon_metabol = model.decode_m(z)
recon_transcript = model.decode_t(z)

# Example usage (R)
library(my_project)

demo <- example_function()
print(demo)

Usage

This tool is run as a set of scripts from the command line. The primary control method of training the VAE in this package is a set of config files in config/. See config/default.yml for the comprehensive set of model and training parameters and what they may look like. Any specified subset of these parameters can be overwritten by providing a custom config. For example:

python scripts/run.py -c config/test.yml

The above code will train a VAE for a single epoch on a random 1% subset of the data used to train the full model. For the standard settings of a full VAE training run:

python scripts/run.py -c config/control.yml

# More usage examples (R)
library(demoProject)

demo <- advanced_function(parameter1 = "value1")
print(demo)

Contribute

Contributions are welcome! If you'd like to contribute, please open an issue or submit a pull request. See the contribution guidelines for more information.

Support

If you have any issues or need help, please open an issue or contact the project maintainers.

License

This project is licensed under the MIT License.