Key concepts and tools

  • GEO accession numbers, EMBL-ENA download portal
  • FTP links, wget for large file transfer
  • bamtofastq (10X Genomics) — BAM to FASTQ conversion
  • Cell Ranger count — alignment and UMI counting
  • 10X Genomics library structure: CB (cell barcode), UMI tags
  • Counts matrix: filtered_feature_bc_matrix/
  • Samplesheet design for a download-and-process pipeline
  • Building a Nextflow pipeline from scratch (no scaffold)

The starting point for the single-cell unit. Published 10X single-cell RNA-seq studies often deposit raw data as aligned BAM files rather than FASTQs, which means you need to reverse the alignment before re-running the standard pipeline. You will locate the appropriate files on EMBL-ENA, build a samplesheet with FTP links, and write a Nextflow pipeline (from an empty repo) that downloads the BAMs, converts them to FASTQs with bamtofastq, and runs cellranger count to produce the counts matrix used in the next two labs.

We are going to recreate findings from this paper, which characterizes isoform expression across cell types in the mouse hippocampus at postnatal day 7 using single-cell RNA-seq.

Background

With 10X Chromium single-cell RNA-seq, many studies deposit data as aligned BAM files that carry cell barcode (CB) and UMI (UR) tags. To regenerate a counts matrix — for example, aligned to a newer reference — you must first reverse the alignment back to FASTQ and then re-run the full Cell Ranger pipeline.

Starting from an empty repo

The GitHub Classroom repo is empty. You need to create the minimal Nextflow structure yourself:

your-repo/
├── main.nf
├── nextflow.config
└── modules/
    ├── download/main.nf
    ├── bamtofastq/main.nf
    └── cellranger_count/main.nf

Finding and downloading the data

  1. Find the GEO accession in the paper and search for it on EMBL-ENA
  2. Locate the FTP links for the BAM files
  3. Encode the sample name and FTP link in a samplesheet.csv

Start with the provided subset samplesheet which points to small subsetted files. Switch to the full samplesheet once the pipeline is working.

Modules to write

Download — Input: tuple val(name), val(ftp_link). Use wget to download the BAM.

bamtofastq — Input: tuple val(name), path(bam). Use the 10X bamtofastq utility. More threads will significantly reduce runtime.

Container: ghcr.io/bf528/cellranger:latest

cellranger count — Input: FASTQs + reference genome path. Point to the FASTQ directory, sample name, and reference. Output: the filtered_feature_bc_matrix/ directory.

Output

A successful run produces a filtered_feature_bc_matrix/ directory per sample containing:

  • matrix.mtx.gz — sparse count matrix
  • features.tsv.gz — gene names
  • barcodes.tsv.gz — cell barcodes

These are the input for QC and analysis in the next lab.

Transferring your repo after the semester

git clone git@github.com:bf528/your-repo.git
cd your-repo/
git remote set-url origin git@github.com:<your_username>/new_repo.git
git push -u origin main

This preserves your full commit history.

Updated: