scVAE: Single-cell variational auto-encoders

scVAE is a command-line tool for modelling single-cell transcript counts using variational auto-encoders.

Install scVAE using pip for Python 3.6 and 3.7:

$ python3 -m pip install scvae

scVAE can then be used to train a variational auto-encoder on a data set of single-cell transcript counts:

$ scvae train transcript_counts.tsv

And the resulting model can be evaluated on the same data set:

$ scvae evaluate transcript_counts.tsv

For more details, see the documentation, which include a user guide and a short tutorial.

Release History

2.1.4 (2020-06-30)

• Better handling of indefinite losses during training.

2.1.3 (2020-06-29)

• Fix loading cell and gene names for H5 data sets.
• Report expected model directory path when scVAE cannot find a model during evaluation for easier troubleshooting.

2.1.2 (2020-04-07)

• Export of decomposition of data sets and latent values as compressed TSV files.
• Export of predictions as compressed TSV files.
• Fix potential crash during t-SNE decomposition.

2.1.1 (2020-02-24)

• Requires TensorFlow 1.15.2 because of a security vulnerability.
• Export of latent values as compressed TSV files.
• Make folder names and filenames more safe on Windows.
• Regrouped analyses, so fewer analyses are performed by default. All available analyses can be performed using --included-analyses all.
• Fix loading of KL divergences when evaluating VAE models.
• Fix crash during model analyses, if the model did not exist.

2.1.0 (2019-11-12)

• Requires Python 3.6 or 3.7 as well as TensorFlow 1.15.
• Documentation with user guide and tutorial.
• Support for sparse matrices in HDF5 format.
• Improved support for Loom files by following conventions.
• Scatter plots of classes against the primary latent feature as well as the two primary latent features against each other when evaluating a model.
• Fix crash related to argparse when using Python 3.6.

2.0.0 (2019-05-18)

• Complete refactor and clean-up including structuring as Python package.
• Easier loading of custom data sets.
• Batch correction included in models for data sets with batch indices.
• Learnable mixture coefficients for the GMVAE model.
• Full covariance matrix for the GMVAE model.
• Sampling from models.

1.0 (2018-04-25)

Initial release.