Full Talapas RNA-seq Run — From Raw FASTQs

Important

This is optional, self-paced material — beyond the three-day workshop. The in-class HPC track (Tutorial 10 — Analysis Pipeline) runs the same ifnb dataset as the laptop tutorials and needs no raw reads. This page is for when you have your own raw 10x FASTQs (or want to reproduce a public dataset such as the 20k NSCLC DTC sample) and need the entire pipeline — from sequencer output all the way to differential abundance — on UO’s Talapas cluster.

This is a start-to-end roadmap: raw FASTQs → a counts matrix with Cell Ranger → the full analysis pipeline (the same numbered scripts as the in-class Talapas track), plus the bonus tracks. It is a first-pass guide that we’ll keep refining; everything it references already exists in the workshop materials.

The roadmap at a glance

Stage What you do Material
0. Cluster basics Log in, modules, sbatch, interactive vs batch jobs Tutorial 09 — SLURM Basics
1. Raw reads → counts cellranger mkfastq / count: BCL → FASTQ → filtered_feature_bc_matrix/ Tutorial 11 — Cell Ranger · cellranger_RNA_count.sbatch
2. Counts → analysis Run the numbered pipeline scripts (01–08), pointing step 01 at your matrix Tutorial 10 — Analysis Pipeline
3. Bonus tracks WGCNA, trajectory, scATAC, spatial, FAIR (standalone scripts 13–17) Tutorial 10 — Bonus tracks

Stage 0 — SLURM basics

If you’ve never submitted a cluster job, start with Tutorial 09 — SLURM Basics: logging in, module load, interactive sessions, and sbatch. Chapter 9 — VS Code & SLURM Basics on Talapas is the deeper reference.

Stage 1 — Raw FASTQs → counts matrix (Cell Ranger)

Tutorial 11 — Raw Data Download + Cell Ranger count walks the whole BCL → FASTQ → counts path on SLURM:

  • how a 10x run goes from BCL → FASTQ (cellranger mkfastq) and FASTQ → counts (cellranger count)
  • installing Cell Ranger and the human reference on Talapas
  • a batch SLURM script — cellranger_RNA_count.sbatch — and scaling to many samples with a SLURM array job
  • inspecting the outputs: the filtered_feature_bc_matrix/ directory that Read10X() consumes

The end product of this stage is a filtered_feature_bc_matrix/ (barcodes / features / matrix) — exactly what the downstream pipeline reads.

Stage 2 — Counts matrix → full analysis pipeline

From here the analysis is identical to the in-class Talapas track — the only change is the data source in step 01. The in-class scripts load ifnb from Bioconductor; to run on your own Cell Ranger output, swap the loader in 01_qc_preprocessing.R for Read10X():

# Replace the muscData loader in 01_qc_preprocessing.R with your matrix:
DATA_DIR <- Sys.getenv("DATA_DIR", "../data/filtered_feature_bc_matrix/")
counts   <- Read10X(data.dir = DATA_DIR)
if (is.list(counts)) counts <- counts[["Gene Expression"]]
seu <- CreateSeuratObject(counts, project = "myproject",
                          min.cells = 3, min.features = 200)
# ...then continue with the same QC / normalize / HVG / scale steps.

Everything downstream — 02–08 — runs unchanged, reading and writing .rds/.csv hand-off files in ../data/. See Tutorial 10 — Analysis Pipeline for the script-by-script table, the run_rscript.sbatch wrapper, and submission commands.

Note

A note on study design. Real differential expression and abundance need genuine biological replicates and a real condition variable. A single dissociated-tumor run (e.g. one 20k NSCLC sample) has neither, so DE/DA on it only exercises the machinery. The in-class ifnb pipeline uses a real CTRL-vs-STIM design across 8 donors precisely so the results are meaningful — keep that in mind when adapting the scripts to your own data.

Stage 3 — Bonus tracks

The standalone scripts 13–17 (WGCNA, trajectory & cell–cell communication, scATAC-seq, spatial, FAIR metadata) parallel the downstream bonus laptop tutorials and are documented in the same Tutorial 10 — Analysis Pipeline page. Scripts 13 (WGCNA) and 15 (scATAC) use their own datasets, 16 (spatial) uses stxBrain, and 14/17 run on the annotated object the core pipeline produces.

For raw scATAC-seq reads specifically (the cellranger-atac count analogue of Stage 1), see the bonus Tutorial 12 — Raw scATAC-seq with cellranger-atac.

Datasets

The bonus Cell Ranger modules use small public PBMC 1k test FASTQs and the GRCh38 reference bundles — download URLs, directory layout, and disk budget are on the Workshop Datasets → Modules 11–12 page. To reproduce this roadmap on a larger, multi-donor sample, the 20k NSCLC DTC (3′ v3.1 Multiplex) public dataset is a good choice — get it from its 10x Genomics landing page (accept the click-through once in a browser, then download the filtered_feature_bc_matrix). Place any downloaded data under /projects/<PIRG>/.../data/ on Talapas, with reference bundles in /projects/<PIRG>/refs/.

Back to top