VeloVAE - Variational Mixtures of ODEs for Inferring Cellular Gene Expression Dynamics
The rapid growth of scRNA-seq data has spurred many advances in single-cell analysis. A key problem is to understand how gene expression changes during cell development. Recently, RNA velocity provided a new way to study gene expression from a dynamical system point of view. Because cells are destroyed during measurement, any scRNA-seq data are just one or very few static snapshots of mRNA count matrices. This makes it hard to learn the gene expression kinetics, as the time information is lost.
VeloVAE is a deep generative model for learning gene expression dynamics from scRNA-seq data. The purpose of our method is to infer latent cell time and velocity simultaneously. This is achieved by using variational Bayesian inference with neural networks. VeloVAE is based on the biochemical process of converting nascent mRNA molecules to mature ones via splicing. The generative model is constrained by a set of ordinary differential equations. VeloVAE is capable of handling more complex gene expression kinetics compared with previous methods.
The package depends on several main-stream packages in computational biology and machine learning, including [scanpy](https://scanpy.readthedocs.io/en/stable/), [PyTorch](https://pytorch.org/), [scikit-learn](https://scikit-learn.org/stable/). We suggest using a GPU to accelerate the training process.
A sample jupyter notebook is available [here](notebooks/velovae_example.ipynb). Notice that the dataset in the example is from [scVelo](https://scvelo.readthedocs.io/), so you would need to install scVelo.
Raw data
{
"_id": null,
"home_page": "https://github.com/welch-lab/VeloVAE",
"name": "velovae",
"maintainer": "",
"docs_url": null,
"requires_python": "",
"maintainer_email": "",
"keywords": "RNA velocity,single-cell RNA,variational autoencoder,deep generative Model,bayesian inference,ordinary differential equation",
"author": "Yichen Gu",
"author_email": "gyichen@umich.edu",
"download_url": "https://files.pythonhosted.org/packages/a3/c8/e0b0604419fe1dde6a74aaf3462e7ec1d99c23e8a7f8c38d6ad3a8c4d128/velovae-0.1.2.tar.gz",
"platform": null,
"description": "VeloVAE - Variational Mixtures of ODEs for Inferring Cellular Gene Expression Dynamics\n\nThe rapid growth of scRNA-seq data has spurred many advances in single-cell analysis. A key problem is to understand how gene expression changes during cell development. Recently, RNA velocity provided a new way to study gene expression from a dynamical system point of view. Because cells are destroyed during measurement, any scRNA-seq data are just one or very few static snapshots of mRNA count matrices. This makes it hard to learn the gene expression kinetics, as the time information is lost.\n\nVeloVAE is a deep generative model for learning gene expression dynamics from scRNA-seq data. The purpose of our method is to infer latent cell time and velocity simultaneously. This is achieved by using variational Bayesian inference with neural networks. VeloVAE is based on the biochemical process of converting nascent mRNA molecules to mature ones via splicing. The generative model is constrained by a set of ordinary differential equations. VeloVAE is capable of handling more complex gene expression kinetics compared with previous methods.\n\nThe package depends on several main-stream packages in computational biology and machine learning, including [scanpy](https://scanpy.readthedocs.io/en/stable/), [PyTorch](https://pytorch.org/), [scikit-learn](https://scikit-learn.org/stable/). We suggest using a GPU to accelerate the training process.\n\nA sample jupyter notebook is available [here](notebooks/velovae_example.ipynb). Notice that the dataset in the example is from [scVelo](https://scvelo.readthedocs.io/), so you would need to install scVelo.\n",
"bugtrack_url": null,
"license": "BSD 3-Clause License",
"summary": "Bayesian Inference of RNA Velocity",
"version": "0.1.2",
"project_urls": {
"Homepage": "https://github.com/welch-lab/VeloVAE"
},
"split_keywords": [
"rna velocity",
"single-cell rna",
"variational autoencoder",
"deep generative model",
"bayesian inference",
"ordinary differential equation"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "50df6b41797929dc39dba8f5204cf509fbb396c7850f99c274bdba48b1741ddd",
"md5": "f520218b8d894a3d5cc9d8876e0f8d4e",
"sha256": "6c582254e39fa80d514d0ec45f02d3e1d5c9659426d8684d24f434303cd717ea"
},
"downloads": -1,
"filename": "velovae-0.1.2-py3-none-any.whl",
"has_sig": false,
"md5_digest": "f520218b8d894a3d5cc9d8876e0f8d4e",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 3532378,
"upload_time": "2023-05-24T04:26:59",
"upload_time_iso_8601": "2023-05-24T04:26:59.254150Z",
"url": "https://files.pythonhosted.org/packages/50/df/6b41797929dc39dba8f5204cf509fbb396c7850f99c274bdba48b1741ddd/velovae-0.1.2-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": "",
"digests": {
"blake2b_256": "a3c8e0b0604419fe1dde6a74aaf3462e7ec1d99c23e8a7f8c38d6ad3a8c4d128",
"md5": "8d3d75c149b5554ebffc39b39f6af9f9",
"sha256": "30b41fc3c5806649ad645697f20918a3e9f3e8ae589047469ad43b27b0532d68"
},
"downloads": -1,
"filename": "velovae-0.1.2.tar.gz",
"has_sig": false,
"md5_digest": "8d3d75c149b5554ebffc39b39f6af9f9",
"packagetype": "sdist",
"python_version": "source",
"requires_python": null,
"size": 3077420,
"upload_time": "2023-05-24T04:27:02",
"upload_time_iso_8601": "2023-05-24T04:27:02.487790Z",
"url": "https://files.pythonhosted.org/packages/a3/c8/e0b0604419fe1dde6a74aaf3462e7ec1d99c23e8a7f8c38d6ad3a8c4d128/velovae-0.1.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2023-05-24 04:27:02",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "welch-lab",
"github_project": "VeloVAE",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "velovae"
}
JSON
