Awesome

VelvetVAE: deep dynamical modelling from temporal transcriptomics.

Analysis software for modelling dynamics of developing systems as neural stochastic differential equation (nSDE) systems using deep generative modelling and time-resolved scRNA-seq data (such as metabolic labelling).

For more details see the pre-print: Maizels et al, 2023.

This package is under active development: it will soon be available on pip with documentation and tutorials.

Installation

Currently, velvetVAE + velvetSDE can be installed as follows:

pip install git+https://github.com/rorymaizels/velvetVAE --user

If you encounter dependency issues, it might be a good idea to set up a virtual environment for velvet:

module load Python/3.10.8-GCCcore-12.2.0 #ensure you have the right python installed
python -m venv .pyenv
source .pyenv/bin/activate
pip install --upgrade pip
pip install --no-cache-dir git+https://github.com/rorymaizels/velvetVAE

# test installation
python
import velvetvae

Usage

Until tutorials and documentation are completed, demonstration of how to use velvet can be found in the code reproducing the analysis from Maizels et al, 2023, available here. In particular, useful notebooks may include:

• Basic visualisation of velocity dynamics using velvet
• Benchmarking code with standard pipeline for velvet
• Benchmarking code with standard pipeline for svelvet (with splicing data)
• Benchmarking code with standard pipeline for velvetSDE
• Downstream analysis looking at cell fates

For details on raw data processing for sci-FATE/sci-FATE2 using dynast, see: https://github.com/rorymaizels/sciFATE2_processing

For any questions, either raise an issue or email rory.maizels@crick.ac.uk.