RNA-seq Workflow

This project aims to build a simple, reproducible RNA-seq processing workflow.

Task

Process raw RNA-seq sequencing reads into quality-controlled, analysis-ready gene-level count data, with summary quality-control information produced alongside the results.

Constraints

• The workflow should be reproducible.
• It should support typical RNA-seq input data.
• It should produce gene-level count data suitable for downstream analysis.
• Quality-control information should be generated and summarized.
• The workflow should remain reasonably simple and suitable for local execution.
• Differential expression analysis is outside the scope of the initial version.

Test data

test-data/fastq/SRR6357070_{1,2}.fastq.gz is a small paired-end S. cerevisiae RNA-seq sample (50,000 read pairs, ~4.3MB total) used to build and exercise a first, simple version of the workflow. It's the WT_REP1 sample from nf-core's official rnaseq test dataset (original source: GEO GSE110004).

test-data/reference/transcriptome.fasta and test-data/reference/genes.gtf are the matching transcriptome/annotation fixture from the same nf-core test-datasets repo, used by Salmon and the gene-counting step. Note: this is fixture data for testing the workflow, not the reference you'd use for a real analysis.

The gene-counting step (transcript→gene mapping) assumes a standard GTF with transcript_id/gene_id attributes (e.g. Ensembl-style), which this fixture GTF has. A GFF3 or nonstandard GTF may need extra options — see the comment on the tx2gene rule in the Snakefile.

This fixture is used by config.test.yaml, the config for testing the pipeline itself. To run against a real dataset, create your own config file (see "Running the workflow" below) pointing at your own samples and reference — it isn't tied to the pipeline.

The files are checked into the repo (small enough, ~350KB) so the workflow is reproducible without a network dependency. To re-fetch them if needed:

mkdir -p test-data/fastq test-data/reference
curl -sL -o test-data/fastq/SRR6357070_1.fastq.gz "https://raw.githubusercontent.com/nf-core/test-datasets/rnaseq/testdata/GSE110004/SRR6357070_1.fastq.gz"
curl -sL -o test-data/fastq/SRR6357070_2.fastq.gz "https://raw.githubusercontent.com/nf-core/test-datasets/rnaseq/testdata/GSE110004/SRR6357070_2.fastq.gz"
curl -sL -o test-data/reference/transcriptome.fasta "https://raw.githubusercontent.com/nf-core/test-datasets/rnaseq/reference/transcriptome.fasta"
curl -sL -o test-data/reference/genes.gtf "https://raw.githubusercontent.com/nf-core/test-datasets/rnaseq/reference/genes.gtf"

Running the workflow

Set up the environment once:

conda env create -f environment.yml
conda activate rnaseq-workflow

Then run the pipeline, always passing an explicit --configfile (there's no default — see below) and adjusting --cores to what's available:

snakemake --configfile config.test.yaml --cores 4

For a real dataset, copy config.test.yaml to your own config file and point it at your samples/reference — give it its own results_dir (e.g. results/<name>/) so different runs never mix or overwrite each other's outputs. Within a config's results_dir:

• counts/gene_counts.csv — the gene-level count matrix.
• qc/multiqc_report.html — aggregated QC (FastQC, fastp, Salmon).

Snakemake merges (rather than replaces) a --configfile with any default set via a configfile: directive in the Snakefile — which is why there isn't one; passing --configfile is required, and omitting it fails fast with a clear error rather than silently mixing two configs' samples together.

The pipeline: FastQC (raw QC) → fastp (trimming) → Salmon (pseudo-alignment + transcript quantification) → pytximport (gene-level summarization) → MultiQC (QC summary).