Skip to content

This project uses Variational Autoencoders (VAEs) to identify key drivers of cancer metastasis by encoding high-dimensional gene expression data into a low-dimensional space. Combining constrained optimization and co-occurrence network analysis, it uncovers critical genes and gene pairs for potential biomarkers and targeted therapies.

Notifications You must be signed in to change notification settings

meghana1090/VAE-Cancer-Metastasis

Repository files navigation

Project Overview

This project leverages a Variational Autoencoder (VAE) to identify key drivers of cancer metastasis by encoding high-dimensional gene expression data into a low-dimensional space. By combining constrained optimization and co-occurrence network analysis, the study finds critical genes and gene pairs that could serve as potential biomarkers or targets for therapy.

This repository includes all of the datasets used in the study, along with code for the model(depmap_vae.ipynb), network analysis (data_analysis.ipynb), and an optimization script designed to run on a high-performance computing server (optimization_script.py).

Due to size limitations, large files such as the VAE code and gene perturbation data are stored externally. You can access them below:

(https://drive.google.com/drive/folders/1W__Ik7l4dRb_fhgTD0fa2HBkIpNIYBz5?usp=drive_link)

About

This project uses Variational Autoencoders (VAEs) to identify key drivers of cancer metastasis by encoding high-dimensional gene expression data into a low-dimensional space. Combining constrained optimization and co-occurrence network analysis, it uncovers critical genes and gene pairs for potential biomarkers and targeted therapies.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published