scarf 0.9.0

Scarf

Scarf is a Python package that performs memory-efficient analysis of single-cell genomics data.

• Analyze atlas scale datasets on your laptop (tested with up to 4 million cells)

• Perform analysis of scATAC-Seq data (datasets with up to 700K cells with 1 million peaks tested) under 10 GB RAM

• Make parallel implementations of UMAP and tSNE (SG-tSNE) for quick cell embedding

• Perform hierarchical clustering that gives interpretable cluster relationships

• Sub-sample highly representative cells using state-of-the-art TopACeDo method

• Perform quick and accurate projections of cells from one dataset onto another or integrate multiple datasets.

Preprint describing Scarf is out on Biorxiv

Install Scarf with:

pip install scarf

Read the documentation here: scarf.rtfd.io or jump to a basic workflow of Scarf

Known issues:

DataStore.run_tsne() does not work on Windows and Mac: We currently ship a pre-compiled version SG-tSNE. This will be fixed in future updates.

High memory consumption: If you are using a version of Scarf less than 0.7.0 and have dask version >2021.03.1 then you might face high memory consumption issues. The solution is to install dask==2021.03.1. The latest version of Scarf automatically solves this issue.

Scarf was recently renamed to scarf from scarf-toolkit on PyPi. So if you previously installed using pip install scarf-toolkit then you need to first uninstall using pip uninstall scarf-toolkit. Thereafter you can install the recent version of Scarf using pip install -U scarf