Installation Guide

You can easily install CytofDR with just one command! This allows you to perform many DR methods as a one-stop solution. Follow the guide here to get started!

Conda

We are officially on conda!! This is actually our recommended way of installing and running CytofDR. To install, simply run the following:

conda install -c kevin931 cytofdr -c conda-forge -c bioconda

If you need to learn more about how to create and manage conda environments, you can take a look at their documentation.

PyPI

Our package is also on PyPI, which you can easily install with the following command:

pip install CytofDR

And voila, that’s it!

Core Dependencies

As an omnibus package, we naturally require lots of dependencies. However, due to inconsistencies with packaging, we only require some core dependencies to be installed. This is the easiest for users. We also list some additional dependencies that need manual installation and care to get working! We will walk you through both processes!

The core dependencies are required for CytofDR. They should be automatically installed with pip or conda processes list above, but if there is an issue, you can elect to install them on your own.

  • scikit-learn

  • numpy

  • scipy

  • umap-learn

  • openTSNE

  • phate

  • annoy

  • matplotlib

  • seaborn

Optional Dependencies

There are some optional dependencies for additional DR methods that we can support. They will not affect other core methods. If you want to use them and integrate them into this package, follow each of the guides below individually, and be sure to check the links to the original repositories and guides for intsllation.

ZIFA

Zero-Inflated Factor Analysis (ZIFA) can be easily installed from this repository by the original authors (Pierson & Yau, 2015). This package is compatible with our core dependencies. To install,

git clone https://github.com/epierson9/ZIFA
cd ZIFA
python setup.py install

Then, you will be able to use ZIFA with CytofDR.

SAUCIE

Although SAUCIE performs quite well, it does not have compatibility with our core dependencies. Some care is needed to install from source. The repository is not currently packaged (nor does it have a proper open-source license for us to do be able to do anything). To install, first, you will need to have python 3.7 and the core dependencies along with CytofDR installed in your environment. For this, we highly recommend using conda to manage this enviroment. Then, you will need to install the following:

conda activate your_environment
conda install tensorflow=1.15
conda install -c bioconda fcsparser
pip install fcswrite

Then, since SAUCIE is not actually installable, you will need to place it in your working directory to make it run:

git clone https://github.com/KrishnaswamyLab/SAUCIE

These steps should allow you to use SAUCIE as intended. Of course, you can use pip if you prefer.

Note

Only tensoflow 1.x is supported. This may cause issues with other dependencies and python version of CytofDR in the future.

Warning

SAUCIE has a known issue of being able to run only once after import using CytofDR. We don’t yet have a workaround for this.

GrandPrix

GrandPrix can be installed from source with the original authors’ GitHub repository (Ahmed et al., 2019). Again, you will need python 3.7 and tensorflow 1.x to get this working. To install, you can simply use the following:

conda activate your_environment
conda install tensorflow=1.15
conda install -c conda-forge gpflow

git clone https://github.com/ManchesterBioinference/GrandPrix
cd GrandPrix
python setup.py install

This should be compatible with SAUCIE in the same environment.

Note

The original authors recommend installing GPflow from source. We recommend installing from a pip or conda for easier installation.