polyphony

This repository contains the backend (model, web server, data manager) for Polyphony, our interactive transfer-learning framework for reference-based single-cell data analysis.

Install

Install from source

git clone https://github.com/ChengFR/polyphony.git
make install

To install the package for development, run:

conda env create -f environment.yml
conda activate polyphony-env

make install-develop

Usage

Run the example

Start the server by

polyphony

See polyphony-vis for frontend installation and usage instructions.

Use your own dataset

Polyphony supports using your own dataset in the format of AnnData.

from polyphony import Polyphony
from polyphony.data import QryAnnDataManager, RefAnnDataManager

ref_dataset = RefAnnDataManager(ref_adata, {'batch_key': batch_key, 'cell_type_key': cell_type_key})
qry_dataset = QryAnnDataManager(qry_adata, {'batch_key': batch_key, 'pred_key': pred_key})

pp = Polyphony('exp', ref_dataset, qry_dataset)

When building the Reference or the Query Dataset for Polyphony, you need to specify the following key names.

• batch_key: the name of the batch name field in adata.obs
• cell_type_key: the name of the cell type field in adata.obs
• pred_key: the name of a reserved field in adata.obs for cell type predictions

See 2. Load Dataset for the full example.

Currently, using external datasets is not supported in the web application.

Citation

To cite Polyphony in your work, please use:

@article{cheng2022polyphony,
  title = {Polyphony: an {Interactive} {Transfer} {Learning} {Framework} for {Single}-{Cell} {Data} {Analysis}},
  author = {Cheng, Furui and Keller, Mark S. and Qu, Huamin and Gehlenborg, Nils and Wang, Qianwen},
  journal = {OSF Preprints},
  year = {2022},
  month = apr,
  doi = {10.31219/osf.io/b76nt},
  url = {https://osf.io/b76nt/},
  language = {en}
}