SISUA

|SISUA_design|

.. |SISUA_design| image:: https://drive.google.com/uc?export=view&id=1PvvG61_Rgbv_rqT6sCeb1XB6CtdiCMXX :width: 405 :height: 249

Semi-supervised Single-cell modeling:

• Free software: MIT license
• Documentation: https://github.com/trungnt13/sisua/tree/master/docs.

Reference:

• Trung Ngo Trong, Roger Kramer, Juha Mehtonen, Gerardo González, Ville Hautamäki, Merja Heinäniemi. "SISUA: SemI-SUpervised Generative Autoencoder for Single Cell Data", ICML Workshop on Computational Biology, 2019. [pdf]__

.. __: https://doi.org/10.1101/631382

Installation

You only need Python 3.6, the stable version of SISUA installed via pip:

pip install sisua

Install the nightly version on github:

pip install git+https://github.com/trungnt13/sisua@master

For developers, we create a conda environment for SISUA contribution sisua_env__

conda env create -f=sisua_env.yml

.. __: https://github.com/trungnt13/sisua/blob/master/sisua_env.yml

Getting started

a. The basics: * Datasets description__ * Models specification * Basic API and work-flow__ b. Single-cell analysis: * Latent space * Imputation of genes expression * Prediction of protein markers c. Advanced technical topics: * Probabilistic embedding__ * Hierarchical modeling (coming soon) * Causal analysis (coming soon) * Cross datasets analysis (coming soon) d. Benchmarks: * Scalability test__ * Fine-tuning networks * Data normalization

.. __: https://github.com/trungnt13/sisua/blob/master/docs/dataset_description.md .. __: https://github.com/trungnt13/sisua/blob/master/tutorials/basics.py .. __: https://github.com/trungnt13/sisua/blob/master/tutorials/probabilistic_embedding.py .. __: https://github.com/trungnt13/sisua/blob/master/tests/scalability.py

Roadmap

1. [x] Multi-OMICs single-cell dataset (link__)
2. [x] Disentanglement VAE for multi-OMICs data (link__)
3. [x] New model: FactorVAE, BetaVAE, MIxture Semi-supervised Autoencoder (MISA) (link__)
4. [ ] Better imputation via hierarchical latents model.
5. [ ] Release SISUA 2

.. __: https://github.com/trungnt13/sisua/blob/master/sisua/data/single_cell_dataset.py .. __: https://github.com/trungnt13/sisua/blob/master/sisua/models/fvae.py .. __: https://github.com/trungnt13/sisua/blob/master/sisua/models/vae.py

Toolkits

We provide binary toolkits for fast and efficient analyzing single-cell datasets:

• sisua-train__: train single-cell modeling algorithms, support training multiple systems in parallel.
• sisua-analyze__: evaluate, compare, and interpret trained model.
• sisua-embed__: probabilistic embedding for semi-supervised training.
• sisua-data__: coming soon

.. __: https://github.com/trungnt13/sisua/blob/master/bin/README.rst .. __: https://github.com/trungnt13/sisua/blob/master/bin/README.rst .. __: https://github.com/trungnt13/sisua/blob/master/bin/README.rst .. __: https://github.com/trungnt13/sisua/blob/master/bin/README.rst

Some important arguments:

-model name of function declared in models__

        - ``scvi``: single-cell Variational Inference model
        - ``dca``: Deep Count Autoencoder
        - ``vae``: single-cell Variational Autoencoder
        - ``movae``: SISUA

-ds name of dataset declared in data__.

        Description of all predefined datasets is in docs__.

        Some good datasets for practicing:

        - ``pbmc8k_ly``
        - ``cortex``
        - ``pbmcecc_ly``
        - ``pbmcscvi``
        - ``pbmcscvae``

.. __: https://github.com/trungnt13/sisua/tree/master/sisua/models .. __: https://github.com/trungnt13/sisua/tree/master/sisua/data .. __: https://github.com/trungnt13/sisua/blob/master/docs/dataset_description.md

Configuration

By default, the data will be saved at your home folder at ~/bio_data, and the experiments' outputs will be stored at ~/bio_log

You can customize these two paths using the environment variables:

• For storing downloaded and preprocessed data: SISUA_DATA
• For the experiments: SISUA_EXP

For example:

.. code-block:: python

import os os.environ['SISUA_DATA'] = '/tmp/bio_data' os.environ['SISUA_EXP'] = '/tmp/bio_log'

from sisua.data import EXP_DIR, DATA_DIR

print(DATA_DIR) # /tmp/bio_data print(EXP_DIR) # /tmp/bio_log

or you could set the variables in advance:

.. code-block:: bash

export SISUA_DATA=/tmp/bio_data export SISUA_EXP=/tmp/bio_log python sisua/train.py

or using the provided toolkit: sisua-train