Semi-supervised Omics Factor Analysis (SOFA)

Introduction

Here we present Semi-supervised Omics Factor Analysis (SOFA), a multi-omics integration method, that incorporates known sources of variation into the model and focuses the latent factor discovery on novel sources of variation. The SOFA method is implemented in Python using the Pyro framework for probabilistic programming.

We are still working on improvements to the SOFA package. Please expect breaking changes. If you find a bug or have ideas how to make the user experience of SOFA smoother please open an issue.

Installation

To install SOFA first create Python 3.8 environment e.g. by

conda create --name sofa-env python=3.8
conda activate sofa-env

and install the package using

pip install biosofa

How to use SOFA for multi-omics analyses

A detailed manual with examples and how to use SOFA can be found here https://tcapraz.github.io/SOFA/index.html.

How to cite SOFA

Semi-supervised Omics Factor Analysis (SOFA) disentangles known sources of variation from latent factors in multi-omics data

Capraz, T., Vöhringer, H.S. and Huber, W.

bioRxiv 2024. doi: 10.1101/2024.10.10.617527.