scAR (single-cell Ambient Remover) is a package for data denoising in single-cell omics.

Overview

scAR

scAR single-cell omics machine learning variational autoencoders denoising

scAR (single cell Ambient Remover) is a package for denoising multiple single cell omics data. It can be used for multiple tasks, such as, sgRNA assignment for scCRISPRseq, identity barcode assignment for cell indexing, protein denoising for CITE-seq, mRNA denoising for scRNAseq, and etc... It is built using probabilistic deep learning, illustrated as follows:

Table of Contents

Installation

Clone this repository,

$ git clone https://github.com/Novartis/scAR.git

Enter the cloned directory:

$ cd scAR

To install the dependencies, create a conda environment:

Please use scAR-gpu if you have an nvidia graphis card and the corresponging driver installed.

$ conda env create -f scAR-gpu.yml

or

Please use scAR-cpu if you don't have a graphis card availalble.

$ conda env create -f scAR-cpu.yml

To activate the scAR conda environment run:

$ conda activate scAR

Usage

There are two ways to run scAR.

  1. Use scAR API if you are Python users
>>> from scAR import model
>>> scarObj = model(adata.X.to_df(), empty_profile)
>>> scarObj.train()
>>> scarObj.inference()
>>> adata.layers["X_scAR_denoised"] = scarObj.native_counts
>>> adata.obsm["X_scAR_assignment"] = scarObj.feature_assignment  # feature assignment, e.g., sgRNAs, tags, and etc.. Only available in 'cropseq' mode

See the tutorials

  1. Run scAR from the command line
$ scar raw_count_matrix.pickle -t technology -e empty_profile.pickle -o output

raw_count_matrix.pickle, a pickle-formatted raw count matrix (MxN) with cells in rows and features in columns
empty_profile.pickle, a pickle-formatted feature frequencies (Nx1) in empty droplets
technology, a string, either 'scRNAseq' or 'CROPseq' or 'CITEseq'

Use scar --help command to see other optional arguments and parameters.

The output folder contains four (or five) files:

output
├── denoised_counts.pickle		# denoised count matrix
├── expected_noise_ratio.pickle	# estimated noise ratio
├── BayesFactor.pickle			# bayesian factor of ambient contamination
├── expected_native_freq.pickle	# estimated native frequencies
└── assignment.pickle			# feature assignment, e.g., sgRNAs, tags, and etc.. Gernerated under 'cropseq' mode

Dependencies

PyTorch 1.8 Python 3.8.6 torchvision 0.9.0 tqdm 4.62.3 scikit-learn 1.0.1

Resources

License

This project is licensed under the terms of License.
Copyright 2022 Novartis International AG.

Reference

If you use scAR in your research, please consider citing our manuscript,

@article {Sheng2022.01.14.476312,
	author = {Sheng, Caibin and Lopes, Rui and Li, Gang and Schuierer, Sven and Waldt, Annick and Cuttat, Rachel and Dimitrieva, Slavica and Kauffmann, Audrey and Durand, Eric and Galli, Giorgio G and Roma, Guglielmo and de Weck, Antoine},
	title = {Probabilistic modeling of ambient noise in single-cell omics data},
	elocation-id = {2022.01.14.476312},
	year = {2022},
	doi = {10.1101/2022.01.14.476312},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {https://www.biorxiv.org/content/early/2022/01/14/2022.01.14.476312},
	eprint = {https://www.biorxiv.org/content/early/2022/01/14/2022.01.14.476312.full.pdf},
	journal = {bioRxiv}
}
Issues
  • Stochastic rounding to integers for downstream use in TotalVI/SCVI

    Stochastic rounding to integers for downstream use in TotalVI/SCVI

    Hi Caibin,

    I tried using scar's output as input for TotalVI/SCVI. As expected, those gave an error because the input is not integer anymore. I would suggest implementing stochastic rounding to integers as done in SoupX.

    Let me know if you're interested and I can find the time to implement it.

    Regards, Mikhael

    enhancement 
    opened by mdmanurung 9
  • BiocondaBot not triggered

    BiocondaBot not triggered

    Hi @fgypas , I made a new release v0.4.1 but bioconda somehow is not triggered upon the new release.

    In the new release, some codes related to building process have been refactored.

    • All information in setup.py (deleted) is integrated into setup.cfg.
    • An extra pyproject.toml file is added.

    I am wondering whether these affect the bioconda-recipes.

    Many thanks, Caibin

    opened by CaibinSh 7
  • New release

    New release

    Hi @fgypas ,

    I am making a new release. There are mainly three changes: 1) addition of a readthedocs; 2) code reformatting via black and pylint (pylint now can score >7, so I have increase the standard in the Action test from 0.5 to 6); 3) renaming 'scAR' to 'scar'.

    I have a couple of questions regarding whether these changes influence the bioconda recipe.

    • Will renaming package name (scAR) require modification in bioconda PR? All uppercase ('scAR') is changed to lowercase ('scar') in everywhere possible (inc. folder, environment, and etc.) But the repo name may stay as 'scAR' for a while, as renaming repo name requires permission from Nick.

    • Should we exclude the folder of datasets in the conda recipe? In addition, a folder, named 'datasets' contains >100 MBs data is added for the tutorial. Should we exclude it?

    question 
    opened by CaibinSh 3
  • Positive-valued denoising results for ADTs with raw 0 counts

    Positive-valued denoising results for ADTs with raw 0 counts

    Hi scar team!

    Thank you for developing this interesting package. I had a question about the resulting denoised values for CITE-seq experiments.

    I've noticed that some cells that originally have a 0 value for an ADT (as a raw count) will have a positive value (>0) for that ADT after the denoising procedure. Below, I show this case for the CD25 ADT in the 10xPBMC5k CITE-seq dataset (from the tutorial at https://scar-tutorials.readthedocs.io/en/latest/tutorials/scAR_tutorial_denoising_CITEseq.html).

    I'm a bit confused about how to best interpret these values and how they are occurring. Should these be set to 0 after the denoising procedure?

    Screen Shot 2022-05-25 at 1 16 37 AM question 
    opened by diegoalexespi 2
  • Number of training epochs + batch size

    Number of training epochs + batch size

    Dear scAR-Team,

    thank you for developing this package. I am currently exploring it and I would like to ask you

    1. how do you determine the number of epochs the user should use for feature_type = "mRNA"? In your tutorials you used 400 epochs and in your paper you mentioned that you fixed the epochs to 800. I applied it for various batch sizes (up to 1000) and noticed that the model is sensitive to it.

    2. I noticed that you use rather small batch-size - is scAR sensitive to the batch-size, it is just due to computational limitations or due to better perfromance?

    Thank you in advance!

    Best,

    question 
    opened by KalinNonchev 1
  • bump to version 0.3.2

    bump to version 0.3.2

    fix(*): changelog docs: adding docstring in documentation docs: adding Release notes in documentation docs: adding docstring in documentation test: adding semantic release refactor: further refactoring codes fix semantic release

    opened by CaibinSh 1
  • ask for permission of Webhooks

    ask for permission of Webhooks

    Hi @kliatsko ,

    We are currently refactoring and adding functionalities to scAR.

    Could you please grant the Webhooks permission for us to automate the documentation?

    Many thanks in advance. Best regards, Caibin on behalf of the scar team @fgypas @Tobias-Ternent @mr-nvs @AlexMTYZ.

    help wanted 
    opened by CaibinSh 1
  • New release

    New release

    • Additions of readthedocs
    • Code refactoring
    1. Renaming module names, e.g. changing "scAR" -> "scar"
    2. Renaming parameter names, e.g.

    changing "scRNAseq_tech" -> "feature_type" changing "model" -> "count_model" changing "scRNAseq_tech" -> "feature_type"

    • Black and Pylint re-formatting the code
    enhancement 
    opened by CaibinSh 1
  • Black github action

    Black github action

    Addition of black github action that runs on every push and every pull request. It shows in the stdout all the changes that need to be made (--diff), but returns exit code 0, even if errors are observed.

    opened by fgypas 1
  • ci(inference): round the counts using stochastic rounding

    ci(inference): round the counts using stochastic rounding

    round the resulting counts (float) for easy interpretation and better integration into other methods, such as totalVI and scVI.

    Credits should go to @mdmanurung

    opened by CaibinSh 0
  • feat(scar.model): addition of a sparsity parameter

    feat(scar.model): addition of a sparsity parameter

    • add a sparsity parameter to control the sparsity of estimated native signals.
    • add a functional testing
    • update examples in scar.model and scar.data_generator
    opened by CaibinSh 0
Releases(v0.4.3)
  • v0.4.3(Jun 15, 2022)

    Fix

    • setup: Fix a bug to allow sample reasonable numbers of droplets (ef6f7e4)
    • main: Fix a bug in main to set default NN number (794ff17)

    Documentation

    • main: Add scanpy as dependency (252a492)

    Performance

    • main: Set a separate batchsize_infer parameter for inference (8727f04)
    • setup: Add an option of random sampling droplets to speed up calculation (ce042dd)
    • setup: Enable manupulate large-scale emptydroplets (15f1840)
    Source code(tar.gz)
    Source code(zip)
  • v0.4.2(Jun 7, 2022)

  • v0.4.1(May 19, 2022)

    What's Changed

    Feature

    • inference: add a round_to_int parameter to round the counts (float) for easy interpretation and better integration into other methods (#47) (902a2b9) (8694239)

    Build

    • setup: replace setup.py with setup.cfg and pyproject.toml (#51) (3dc999a)

    Chore

    Documentation

    • readthedocs: add scAR_logo image (#51) (c34f362)
    • tutorials: add ci=None to speed up plotting (#51) (902a2b9)

    Contributor

    @CaibinSh and @mdmanurung

    Full Changelog: https://github.com/Novartis/scar/compare/v0.4.0...v0.4.1

    Source code(tar.gz)
    Source code(zip)
  • v0.4.0(May 5, 2022)

  • v0.3.5(May 3, 2022)

  • v0.3.4(May 1, 2022)

  • v0.3.3(May 1, 2022)

  • v0.3.1(Apr 29, 2022)

  • v0.3.0(Apr 27, 2022)

    What's Changed

    Renaming module names, e.g. changing "scAR" -> "scar" Renaming parameter names, e.g.

    "scRNAseq_tech" -> "feature_type" "model" -> "count_model" "empty_profile" -> "ambient_profile" ...

    • Black and Pylint re-formatting the code
    • New release by @CaibinSh in https://github.com/Novartis/scAR/pull/26

    Contributor

    @CaibinSh @fgypas @mr-nvs @Tobias-Ternent

    Full Changelog: https://github.com/Novartis/scAR/compare/v0.2.3...v0.3.0

    Source code(tar.gz)
    Source code(zip)
  • v0.2.3(Apr 20, 2022)

    • Add integration test
    • Black formating
    • Bump version to 0.2.3

    Contributors: @fgypas , @mr-nvs and @CaibinSh

    What's Changed

    • Develop by @CaibinSh in https://github.com/Novartis/scAR/pull/19

    Full Changelog: https://github.com/Novartis/scAR/compare/v0.2.2...v0.2.3

    Source code(tar.gz)
    Source code(zip)
  • v0.2.2(Apr 4, 2022)

    v0.2.2

    • Remove torchaudio
    • Add test data for integration tests
    • Bump version to 0.2.2

    Contributors: @CaibinSh @fgypas

    What's Changed

    • Remove torchaudio, add test data and bump version to 0.2.2 by @fgypas in https://github.com/Novartis/scAR/pull/15

    Full Changelog: https://github.com/Novartis/scAR/compare/v0.2.1-beta...v0.2.2

    Source code(tar.gz)
    Source code(zip)
  • v0.2.1-beta(Apr 1, 2022)

    • fix a typo in scAR-gpu.yml
    • reorganise init.py files

    Contributor: @CaibinSh

    What's Changed

    • Develop by @CaibinSh in https://github.com/Novartis/scAR/pull/12

    Full Changelog: https://github.com/Novartis/scAR/compare/v0.2.0-beta...v0.2.1-beta

    Source code(tar.gz)
    Source code(zip)
  • v0.2.0-beta(Apr 1, 2022)

    • Support for training of the model with CPUs
    • Addition of two yaml files for CPU/GPU installation
    • Refactor of setup.py and structure of the package
    • Addition of tests with pytest
    • Addition of lint checks
    • Automate build with github actions (install package and run lint checks and pytest)
    • Update documentation
    • Version 0.2.0

    Co-authored-by: @CaibinSh @mr-nvs @Tobias-Ternent @fgypas

    What's Changed

    • 0.2.0-release by @fgypas in https://github.com/Novartis/scAR/pull/11

    Full Changelog: https://github.com/Novartis/scAR/commits/v0.2.0-beta

    Source code(tar.gz)
    Source code(zip)
A scanpy extension to analyse single-cell TCR and BCR data.

Scirpy: A Scanpy extension for analyzing single-cell immune-cell receptor sequencing data Scirpy is a scalable python-toolkit to analyse T cell recept

ICBI 123 Jul 21, 2022
A Parameter-free Deep Embedded Clustering Method for Single-cell RNA-seq Data

A Parameter-free Deep Embedded Clustering Method for Single-cell RNA-seq Data Overview Clustering analysis is widely utilized in single-cell RNA-seque

AI-Biomed @NSCC-gz 3 May 8, 2022
7th place solution of Human Protein Atlas - Single Cell Classification on Kaggle

kaggle-hpa-2021-7th-place-solution Code for 7th place solution of Human Protein Atlas - Single Cell Classification on Kaggle. A description of the met

null 8 Jul 9, 2021
Single Red Blood Cell Hydrodynamic Traps Via the Generative Design

Rbc-traps-generative-design - The generative design for single red clood cell hydrodynamic traps using GEFEST framework

Natural Systems Simulation Lab 4 Jun 16, 2022
Official repository for "Restormer: Efficient Transformer for High-Resolution Image Restoration". SOTA for motion deblurring, image deraining, denoising (Gaussian/real data), and defocus deblurring.

Restormer: Efficient Transformer for High-Resolution Image Restoration Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan,

Syed Waqas Zamir 741 Aug 10, 2022
Practical Blind Denoising via Swin-Conv-UNet and Data Synthesis

Practical Blind Denoising via Swin-Conv-UNet and Data Synthesis [Paper] [Online Demo] The following results are obtained by our SCUNet with purely syn

Kai Zhang 220 Aug 7, 2022
Pseudo-Visual Speech Denoising

Pseudo-Visual Speech Denoising This code is for our paper titled: Visual Speech Enhancement Without A Real Visual Stream published at WACV 2021. Autho

Sindhu 91 Jul 30, 2022
EDCNN: Edge enhancement-based Densely Connected Network with Compound Loss for Low-Dose CT Denoising

EDCNN: Edge enhancement-based Densely Connected Network with Compound Loss for Low-Dose CT Denoising By Tengfei Liang, Yi Jin, Yidong Li, Tao Wang. Th

workingcoder 112 Jul 29, 2022
Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021)

Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation (CVPR 2021, official Pytorch implementatio

Microsoft 222 Aug 4, 2022
HiFi-GAN: High Fidelity Denoising and Dereverberation Based on Speech Deep Features in Adversarial Networks

HiFiGAN Denoiser This is a Unofficial Pytorch implementation of the paper HiFi-GAN: High Fidelity Denoising and Dereverberation Based on Speech Deep F

Rishikesh (ऋषिकेश) 120 Aug 1, 2022
Official implementation of Deep Convolutional Dictionary Learning for Image Denoising.

DCDicL for Image Denoising Hongyi Zheng*, Hongwei Yong*, Lei Zhang, "Deep Convolutional Dictionary Learning for Image Denoising," in CVPR 2021. (* Equ

Z80 77 Jul 1, 2022
The source code of the paper "Understanding Graph Neural Networks from Graph Signal Denoising Perspectives"

GSDN-F and GSDN-EF This repository provides a reference implementation of GSDN-F and GSDN-EF as described in the paper "Understanding Graph Neural Net

Guoji Fu 17 Oct 13, 2021
A denoising diffusion probabilistic model (DDPM) tailored for conditional generation of protein distograms

Denoising Diffusion Probabilistic Model for Proteins Implementation of Denoising Diffusion Probabilistic Model in Pytorch. It is a new approach to gen

Phil Wang 83 Jul 29, 2022
Point Cloud Denoising input segmentation output raw point-cloud valid/clear fog rain de-noised Abstract Lidar sensors are frequently used in environme

Point Cloud Denoising input segmentation output raw point-cloud valid/clear fog rain de-noised Abstract Lidar sensors are frequently used in environme

null 66 Aug 3, 2022
Denoising Diffusion Probabilistic Models

Denoising Diffusion Probabilistic Models This repo contains code for DDPM training. Based on Denoising Diffusion Probabilistic Models, Improved Denois

Alexander Markov 5 May 12, 2022
[NeurIPS 2020] Official repository for the project "Listening to Sound of Silence for Speech Denoising"

Listening to Sounds of Silence for Speech Denoising Introduction This is the repository of the "Listening to Sounds of Silence for Speech Denoising" p

Henry Xu 38 Apr 24, 2022
A two-stage U-Net for high-fidelity denoising of historical recordings

A two-stage U-Net for high-fidelity denoising of historical recordings Official repository of the paper (not submitted yet): E. Moliner and V. Välimäk

Eloi Moliner Juanpere 53 Aug 9, 2022
ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models (ICCV 2021 Oral)

ILVR + ADM This is the implementation of ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models (ICCV 2021 Oral). This repository is h

Jooyoung Choi 148 Aug 8, 2022