Negative multinomial variational auto-encoder
Project description
scMaui is python package that implements a variational auto-encoder for multi-omics data integration. The model is capable of handling variable numbers input and output modalities as well as missing modalities. The model also features a range of log-likelihood implementations for determining the reconstruction loss, including the negative binomial or the negative multinomial model.
The package is freely available under a GNU Lesser General Public License v3 or later (LGPLv3+)
Installation
pip scmaui
Usage
import pkg_resources
from scmaui.data import load_data, SCDataset
from scmaui.utils import get_model_params
from scmaui.ensembles import EnsembleVAE
# get some toy data
toy_data_path = pkg_resources.resource_filename('scmaui', 'resources/gtx.h5ad')
adatas = load_data([toy_data_path], names=['gtx'])
dataset = SCDataset(adatas, losses=['negbinom'])
# create an scMaui model
params = get_model_params(dataset)
ensemble = EnsembleVAE(params=params)
# fit the model
ensemble.fit(dataset, epochs=10)
ensemble.summary()
# obtain latent features
latents, _ = ensemble.encode(dataset)
# obtain an imputation
imputed = ensemble.impute(dataset)
# obtain input feature attributions
selected_cells = latents.index.tolist()[:5] # select first 5 cells
explanation = ensemble.explain(dataset, cellids=selected_cells)
Command-line usage
scMaui offers a command line interface for model fitting. The results are stored in an output directory (-output).
scmaui -data adata.h5ad \ -names gtx \ -output <outputdir> \ -epochs 200 \ -ensemble_size 10 \ -nlatent 15 \ -adversarial label1 label2 \ -conditional covariate1 covariate2
Additional information on available hyper-parameters are available through
scmaui -h
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file scmaui-0.0.3.tar.gz
.
File metadata
- Download URL: scmaui-0.0.3.tar.gz
- Upload date:
- Size: 9.5 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.16
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | bd2000f9a53f3abfdccf2f77fc604a706e852038d02f56853f13899baae0b5b6 |
|
MD5 | fadc1ff6d99483e1e51c5683ad506b84 |
|
BLAKE2b-256 | d97e0dc37aa91062279c62ce60be94a7e26772897893845c25e4be42272f70e6 |
File details
Details for the file scmaui-0.0.3-py2.py3-none-any.whl
.
File metadata
- Download URL: scmaui-0.0.3-py2.py3-none-any.whl
- Upload date:
- Size: 9.6 MB
- Tags: Python 2, Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.16
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 222b2b7c9ed275ea063487a026018de20ebdd1d1de0d37386f83e5c173dac272 |
|
MD5 | 1d375309f549ddde3190673dbea40f09 |
|
BLAKE2b-256 | 5fb413c77c0ff1d3187227b6cc252e08f904bd427d59c545738dc7ebd1653991 |