I’ve packaged a minimal Snakemake + Conda workflow that runs FastQC, Trim Galore, STAR, and featureCounts on about 5M reads per sample in about 45 minutes on a 4-core laptop, with every step versioned for reproducibility. If you’re a wet-lab biologist wanting gene counts from raw FASTQ without wrestling with installs, say the word and I’ll share the repo and a short setup note.
I get better alignment when the STAR index is built with --sjdbOverhang set to read length minus 1; if your runs mix 75/100bp, I just set it to 99 to avoid rebuilding (small tradeoff). Docs: GitHub - alexdobin/STAR: RNA-seq aligner.
@OP I burned a day thinking the pipeline was broken until RSeQC’s infer_experiment.py (RSeQC: An RNA-seq Quality Control Package — RSeQC documentation) told me our data were reverse‑stranded — hardcode featureCounts -s 2 (or -s 0 if unstranded) and you’ll save folks a headache; any chance you add a tiny rule for that?
Quick tip: drop MultiQC (https://multiqc.info) at the end so the FastQC/STAR/featureCounts logs roll up into one report — it’s like a tidy lab notebook for your runs. If you switch featureCounts to -g gene_name for friendlier tables, just remember names aren’t unique, so keep gene_id in the main outputs.
@OP One thing I’ve found: if you want human‑readable symbols in the featureCounts output, set -g “gene_name” with GENCODE; otherwise you’ll get Ensembl IDs by default. If you need stable IDs for downstream comparisons, stick with gene_id — no runtime change on a 4‑core laptop with about 5M reads.