scanpy_recipes

Generalized and aggregated analysis functions for working with Anndata and Scanpy.

Introduction

This package is an extension to the scanpy API tentatively called scanpy_recipes, inspired by the way scanpy references their implementations of previous workflows as recipes.

Aims

There are three main aims of the additional functions of scanpy_recipes:

1. simplify scRNA-seq analysis in JAX's Single Cell Biology Lab.
2. use the AnnData object to capture and retain as much metadata as possible. In theory, I'd like the AnnData object to be a self-contained description of the experiment, analysis, and results.
3. integrate seamlessly with the normal scanpy API and as such only extend functionality.

Specifically, scanpy is normally loaded using import scanpy.api as sc. Instead, this package provides this api object as

from scanpy_recipes.api import sc

This is the same scanpy.api object, does not change any scanpy functionality, and only adds additional functionality.

Workflow

The workflow has 3 components with all three generating their own AnnData object. By default, scanpy will modify AnnData objects inplace. Because we're working in a collaborative environment where multiple users of varying skill levels will be using this framework, I wanted this inplace modification to be crystal clear, especially since finding highly variable genes will remove all genes except the highly variable ones from the default expression matrix (they can still be found in the AnnData.raw).

These three components are:

• Specify sample metadata, input/output directories up from and loading adata_raw
• Iteratively filter and QC data, ultimately yielding adata_qc
• Generate adata_redux through HVG selection, dimensionality reduction, and clustering.

Each of these objects will be saved, and you can restart any step by loading the object output by the previous step.

See [examples/Analysis-Template.ipynb] for an example analysis.

Getting Started

Prerequisites

This library doesn't have too many requirements besides scanpy and the basic Python scientific stack (numpy, pandas, matplotlib).

Installing

Until we hit a major release, installation should be done through pip egg installs:

pip install -e git+https://github.com/TheJacksonLaboratory/scanpy_recipes.git#egg=scanpy_recipes

or

git clone https://github.com/TheJacksonLaboratory/scanpy_recipes.git
cd scanpy_recipes
pip install -e .

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

Authors

• Bill Flynn @wflynny

License

Since this package extends scanpy, this work is under the same BSD-3 license. See LICENSE for more details.