README.md

Workflow overview

This workflow is a best-practice workflow for preprocessing counts from single cell RNA sequencing data. The workflow is built using snakemake and follows the Preprocessing and visualization section of the Single Cell Best Practices.

It consists of the following steps:

1. Convert input data to zarr format.
2. Filter low-quality barcodes with scanpy.
3. Correct for ambient RNA contamination with SoupX.
4. Detect doublets with scDblFinder.
5. Normalize counts with scanpy.

Running the workflow

To configure the workflow run, go through the provided config/config.yaml entry by entry and adjust them where necessary. After an initial run, we recommend going through the quality control plots mentioned there and double-checking that the provided threshold values for filtering make sense.

Input data

To specify the input, provide a config/sample_sheet.tsv file with the following layout:

sample_id	raw_counts_path	format
cellranger_1	../path/to/raw_feature_bc_matrix/	10x_mtx
kallisto_bustools_1	../path/to/adata.h5ad	h5ad
alevin_fry_1	../path/to/quants_mat.mtx	mtx

Here, the columns are:

• sample_id: An arbitrary string identifier of a a sample (or dataset).
• raw_counts_path: The path to a file or folder with the raw counts as determined by another tool, for example CellRanger, kallisto bustools or alevin-fry. We really recommend using the raw counts here (and not any pre-filtered counts), as they are instrumental in the correction for ambient RNA contamination. The workflow filters low-quality barcodes itself, and lets you transparently configure and check the filter thresholds (see the comments in config/config.yaml).
• format: Format of the input data given in column raw_counts_path. Choose a format that scanpy can read, so any suffix in one of the scanpy.read_ functions. See https://scanpy.scverse.org/en/stable/api/io.html or double-check the code at: https://github.com/scverse/scanpy/blob/a656a33b080a5c1f64b01e841daad76f35f5ec5f/src/scanpy/io/_read.py#L44-L61"