Skip to content

daklab/SpliceVI

Repository files navigation

SpliceVI

Multimodal VAE for joint modeling of alternative splicing (PSI) and gene expression from single-cell data. Built on scvi-tools.


Installation

1. Create and activate the environment

conda create -n splicevi-env python=3.12
conda activate splicevi-env

2. Install the package

git clone https://github.com/daklab/SpliceVI.git
cd SpliceVI
pip install -e .

This installs all dependencies automatically from pyproject.toml. For W&B logging, additionally run pip install wandb.

After installation, from splicevi import SPLICEVI works from any script or notebook.

Exact environment: requirements.txt contains a full freeze of the development environment for reproducibility reference, but has machine-specific paths and is not intended as a direct install source.


Repository structure

SpliceVI/
├── src/splicevi/
│   ├── splicevi.py          # SPLICEVI model class for GE+AS (training, inference, DE, DS)
│   ├── splicevae.py         # VAE module for GE+AS (encoder/decoder architecture)
│   ├── partialvae.py        # AS Missingness-Aware Partial VAE module
│   ├── eddisplice.py        # EDDISPLICE model class for single-modality AS-only VAE using PARTIALVAE
│   └── __init__.py          # Package exports
├── train_splicevi.py        # Training entry point
├── eval_splicevi.py         # Evaluation entry point (UMAPs, metrics, imputation)
├── train_splicevi.sh        # SLURM job script for training
├── eval_splicevi.sh         # SLURM job script for evaluation
├── pyproject.toml           # Package configuration
└── requirements.txt         # Full conda environment freeze

Usage

Configure paths and hyperparameters at the top of the shell scripts, then submit:

sbatch train_splicevi.sh
sbatch eval_splicevi.sh

Or run directly:

python train_splicevi.py --help
python eval_splicevi.py --help

Future additions:

  • Add tutorial.ipynb for a walkthrough of model setup, training, and application to other datasets.
  • Add trained models to Hugging Face

Citation

If you use SpliceVI in your work, please cite:

Vaidyanathan S, Isaev K, Zweig A, Knowles DA. Robust Integration of Sparse Single-Cell Alternative Splicing and Gene Expression Data with SpliceVI. bioRxiv 2025.11.26.690853. https://doi.org/10.1101/2025.11.26.690853

@article{splicevi,
  title   = {Robust Integration of Sparse Single-Cell Alternative Splicing and Gene Expression Data with {SpliceVI}},
  author  = {Vaidyanathan, Smriti and Isaev, Keren and Zweig, Aaron and Knowles, David A},
  journal = {bioRxiv},
  year    = {2025},
  doi     = {10.1101/2025.11.26.690853}
}

About

Multimodal VAE for integration of sparse alternative-splicing and gene-expression

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors