Metadata-Version: 2.1
Name: scaespy
Version: 1.2.0
Summary: scAEspy: a tool for autoencoder-based analysis of single-cell RNA sequencing data
Home-page: https://gitlab.com/cvejic-group/scaespy
Author: Andrea Tangherloni
Author-email: andrea.tangherloni@unibg.it
License: LICENSE
Keywords: autoencoder,machine learning,dimensionality reduction,single-cell RNA-data
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3.5
Classifier: Programming Language :: Python :: 3.7
Requires-Python: >=3.5,<3.8
Requires-Dist: NumPy (>=1.18)
Requires-Dist: Pandas (>=0.25)
Requires-Dist: matplotlib (>=3.1)
Requires-Dist: scikit-learn (>=0.21)
Requires-Dist: seaborn (>=0.9)
Requires-Dist: Tensorflow (<=1.15,>=1.12)

scAEspy is a user-friendly and standalone tool that embodies six of the most advanced autoencoders (AEs), easily accessible by setting up only two user-defined parameters (i.e., alpha and lambda), and different loss functions, which are fundamental to deal with the different RNA sequencing platforms.

Specifically, scAEspy contains the following most advanced AEs: Variational AE, Gaussian-mixture VAE (GMVAE), Maximum Mean Discrepancy (MMD) AE, MMDVAE (a combination of MMDAE and VAE), and two novel Gaussian-mixture AEs that we developed, called GMMMD and GMMMDVAE.
GMMMD is a modification of the MMDVAE where more than one Gaussian distribution is used to model different modes and only the MMD loss function is used as a divergence function.
GMMMDVAE is a combination of MMDVAE and GMVAE where both the MMD function and the Kullback–Leibler (KL) divergence function are used.

scAEspy gives easy access to the latent space generated by the selected AE, which can be utilised to perform downstream analysis or to generate synthetic cells.

For further information, please visit https://gitlab.com/cvejic-group/scaespy

