Skip to content
Clustering on Network of Samples
C++ C R Other
Branch: master
Clone or download
Latest commit d23a523 Jul 18, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
R as it says in the changelog, I've changed the explicit '%' to 'percen… Jul 16, 2019
inst/extdata updated cell annotations, re-rendered vignette Mar 27, 2019
man as it says in the changelog, I've changed the explicit '%' to 'percen… Jul 16, 2019
src updated Dockerfile to install from master, use BiocManager, update Bi… Jul 14, 2019
vignettes re-rendered walkthrough using k=30, to reduce the number of initial Jul 14, 2019
.Rbuildignore update Dec 17, 2017
.gitignore Export Conos to ScanPy. Version 1.0.2 Jun 26, 2019 as it says in the changelog, I've changed the explicit '%' to 'percen… Jul 16, 2019 Update Jul 10, 2019
DESCRIPTION as it says in the changelog, I've changed the explicit '%' to 'percen… Jul 16, 2019
NAMESPACE Merge branch 'master' of Jul 2, 2019 Update Jul 18, 2019


Clustering on Network of Samples

  • What is Conos? It's a package to wire together large collections of single-cell RNA-seq datasets. It focuses on uniform mapping of homologous cell types across heterogeneous sample collections. For instance, a collection of dozens of peripheral blood samples from cancer patients, combined with dozens of controls. And perhaps also including samples of a related tissue, such as lymph nodes.

  • How does it work? overview Conos applies one of many error-prone methods to align each pair of samples in a collection, establishing weighted inter-sample cell-to-cell links, creating a global joint graph. Cells of the same type will tend to map to each other across many such pair-wise comparisons, forming cliques, that can be recognized as clusters (graph communities).

  • What does it produce? In essense, Conos will take a large, potentially heterogeneous panel of samples and will produce clustering grouping similar cell subpopulations together in a way that will be robust to inter-sample variation:

  • What are the advantages over existing alignment methods? Conos is robust to heterogeneity of samples within collection, as well as noise. The ability to resolve finer subpopulation structure improves as the size of the panel increases.

  • What do I need to run it? Conos is an R package. Currently, it supports pre-processing (filtering, normalization, etc.) of the individual datasets using pagoda2 or Seurat.


Native installations have been tested in Linux. Normal installation should take <10min.

Native installation

Please make sure devtools package is installed (use install.packages("devtools") to install it if needed). Then install pagoda2 (or Seurat), then install conos:


System dependencies

The dependencies are inherited from pagoda2:

Ubuntu Dependencies

Install system dependencies, example here provided for Ubuntu

sudo apt-get update
sudo apt-get -y install build-essential cmake gsl-bin libgsl0-dev libeigen3-dev libboost-all-dev libssl-dev libcurl4-openssl-dev libssl-dev libcairo2-dev libxt-dev libgtk2.0-dev libcairo2-dev xvfb xauth xfonts-base
Red-Hat-based distributions Dependencies

Was tested on AWS linux and Centos 7

yum install cairo-devel pango-devel libXt-devel openssl-devel gsl-devel boost-devel libcurl-devel

It is possible to install pagoda2 and Conos on OS X, however some users have reported issues with OpenMP configuration. For instructions see pagoda2 readme.

Installing Conos as Docker Container

If your system configuration is making it difficult to install Conos natively, an alternative way to get Conos running is through a docker container.

Note: on OS X, Docker Machine has Memory and CPU limit. To control it, please check instructions either for CLI or for Docker Desktop.

Ready-to-run docker image

The docker distribution has the latest version and also includes the Pagoda2 package. To start a docker container, first install docker on your platform and then start the pagoda container with the following command in the shell:

docker run -p 8787:8787 -e PASSWORD=pass

The first time you run the command it will download several images so make sure that you have fast internet access setup. You can then point your browser to http://localhost:8787/ to get an Rstudio environment with pagoda2 and conos installed (log in using credentials rstudio / pass). Explore the docker --mount option to allow access of the docker image to your local files.

Note: if you already downloaded the docker image and want to update it, please run

docker pull vpetukhov/conos:latest

Building docker image on the fly

If you want to build image by your own, download the Dockerfile (available in this repo under /dockers) and run to following command to build it:

docker build -t conos .

This will create a "conos" docker image on your system (be patient, as the build takes ~30-50min or so). You can then run it using the following command:

docker run -d -p 8787:8787 -e PASSWORD=pass --name conos -it conos

Usage example

Alignment of datasets

Please see Conos tutorial for detailed usage. The overall runtime of the tutorial should be ~5 minutes.

Additional examples: forcing better alignment, integrating RNA-seq and ATAC-seq.

Given a list of individual processed samples (pl), Conos processing can be as simple as this:

# construct conos object, where pl is a list of pagoda2 objects 
con <- Conos$new(pl)

# build graph

# find communities

# plot joint graph

# plot panel with joint clustering results

Integration with ScanPy

For integration with ScanPy you need to save Conos files on disk from R session, than upload these files from Python. See the following tutorials:


If you find this pipeline useful for your research, please consider citing the paper:

Barkas N., Petukhov V., Nikolaeva D., Lozinsky Y., Demharter S., Khodosevich K. & Kharchenko P.V. Joint analysis of heterogeneous single-cell RNA-seq dataset collections. Nat. Methods, (2019). doi:10.1038/s41592-019-0466-z

You can’t perform that action at this time.