scRecouter

A Nextflow pipeline to re-process single-cell RNA-seq data from the Sequence Read Archive.

Workflow

• User provides:
  • A table of samples & associated accessions
    • Alternatively, the pipeline can pull accessions from the scRecounter SQL database
  • Associated files required:
    • A table of barcodes to use for cell barcode and UMI identification
    • A table of STAR index directories to use for mapping
• Pipeline:
  • Load accessions from provided table or SQL database
  • For each accession:
    • Use fastq-dump to download a subset of reads as fastq files from the SRA
    • Determine the "best" STAR parameters by mapping the reads using various parameter combinations
      • Parameters: version of cell barcodes, cell barcode length, UMI length, strand, STAR reference index
      • The STAR parameters are selected based on the fraction of valid barcodes
    • Download all reads with fasterq-dump
      • If download fails, try again with fastq-dump using a max of fallback_max_spots reads (see nextflow.config).
    • Map the reads with STARsolo using the "best" STAR parameters

Manuscript

scBaseCamp: An AI agent-curated, uniformly processed, and continually expanding single cell data repository. Nicholas D Youngblut, Christopher Carpenter, Jaanak Prashar, Chiara Ricci-Tam, Rajesh Ilango, Noam Teyssier, Silvana Konermann, Patrick Hsu, Alexander Dobin, David P Burke, Hani Goodarzi, Yusuf H Roohani. bioRxiv 2025.02.27.640494; doi: https://doi.org/10.1101/2025.02.27.640494

Installation

Conda & mamba install

mamba is needed to run the pipeline. It is a faster version of conda. mamba can be installed via conda. You can use conda instead of mamba if you prefer.

Nextflow install

It is easiest to install Nextflow using mamba (or conda).

mamba create -n nextflow_env -c bioconda nextflow

Make sure to activate the environment before running the pipeline:

mamba activate nextflow_env

All other dependencies will be installed by Nextflow.

Pipeline install

Clone the repo

git clone https://github.com/ArcInstitute/scRecounter.git \
  && cd scRecounter

Pipeline conda environments (if running locally)

The pipeline uses conda environments to manage dependencies. Nextflow will automatically create the environments as long as mamba is installed.

Note: it can take a while to create the environments, even with mamba.

Pipeline Docker containers (if running on GCP)

The pipeline defaults to using custom Docker containers hosted on Google Artifact Registry.

You can build the Docker containers yourself. See ./docker/README.md for details. Be sure to update the profiles.config file to point to the new containers.

Usage

Input

Accessions table

Lists the samples and their associated SRA experiment accessions.

This table is not required if the pipeline is pulling accessions from the scRecounter SQL database. To pull accessions from the database, do not provide --accessions via the command line.

Example:

sample	accession	organism
SRX22716300	SRR27024456	human
SRX25994842	SRR30571763	mouse

organism is optional. It will determine the STAR index to use for mapping. Otherwise all indexes will be used for parameter selection.

Barcode table

Lists all of the possible barcodes that will be used to determine the cell barcode and UMI for the samples.

Example:

name	cell_barcode_length	umi_length	file_path
737K-arc-v1	16	12	/large_storage/goodarzilab/public/scRecount/genomes/737K-arc-v1.txt
737K-august-2016	16	12	/large_storage/goodarzilab/public/scRecount/genomes/737K-august-2016.txt
3M-february-2018	16	10	/large_storage/goodarzilab/public/scRecount/genomes/3M-february-2018.txt

STAR index table

Lists the STAR index files that will be used to map the reads.

Example:

Organism	Star Index Path
human	/large_storage/goodarzilab/public/scRecount/genomes/star_refData_2020_hg38
mouse	/large_storage/goodarzilab/public/scRecount/genomes/star2.7.11_refData_2020_mm10

If organism is provided in the Accessions table, the STAR index will be selected based on the organism column. Thus, it reduces the number of parameter combinations that need to be tested.

Nextflow run

Test runs

Local run with provided accessions:

nextflow run main.nf \
  -work-dir tmp/work \
  -profile conda,trace,report,vm,vm_dev,dev,acc_dev

Local run with provided accessions (problematic datasets)

nextflow run main.nf \
  -work-dir tmp/work \
  -profile conda,trace,report,vm,vm_dev,dev,acc_dev_problems

With conda, accessions pulled from scRecounter database:

nextflow run main.nf \
  -work-dir tmp/work \
  -profile conda,trace,report,vm,vm_dev,dev,no_acc_dev

GCP run with provided accessions:

nextflow run main.nf \
  -profile docker,trace,report,gcp,gcp_dev,dev,acc_dev

GCP run with accessions pulled from scRecounter SQL database:

nextflow run main.nf \
  -profile docker,trace,report,gcp,gcp_dev,dev,no_acc_dev

Characterize datasets

Use just a small subset of reads in the dataset to identify library prep method, species, etc.

nextflow run /home/nickyoungblut/dev/nextflow/scRecounter/main.nf \
  -work-dir gs://arc-ctc-nextflow/scRecounter/work \
  -profile docker,gcp \
  -ansi-log false \
  --max_spots 100000 \
  --output_dir gs://arc-ctc-nextflow/scRecounter/results/ \
  --accessions TMP/SRX22716300.csv

Deploy to GCP Cloud Run

See ./docker/sc-recounter-run/README.md for details.

Contributing

Feel free to fork the repository and submit a pull request. However, the top priority is to keep SRAgent functioning for the ongoing scBaseCamp project.