Quick-start checklist

Tutorial contents · Start the full tutorial: Overview →

Use this page as a navigation aid, not as a substitute for the chapter-level checks.

1. Review the resource-planning companion, then follow the installation and setup chapter, activate the tutorial environment, and verify every required executable and R package before downloading or processing data. Run the synthetic smoke test as a fast known-output check.
2. State the biological comparison, experimental unit, replicate structure, likely nuisance variables, organism, and reference build.
3. Copy the files described in the config/templates/ guide and replace every REVIEW_ME, placeholder path, and example sample.
4. Record checksums for reads and reference resources. Confirm that FASTA, indexes, annotation, chromosome names, and blacklist all use the same build.
5. Run FastQC on raw mates and summarize only the intended reports with MultiQC. Decide whether trimming is justified by the evidence.
6. If trimming, process both mates together and rerun read-level QC.
7. Align paired reads to the verified reference. Record the executable and version, command, exit status, and mapping summary.
8. Apply declared mapping-quality, flag, duplicate, unwanted-contig, and pair-preserving blacklist rules. Validate BAM integrity, pairing, fragment- length periodicity, mitochondrial fraction, and TSS enrichment where the required annotation is suitable.
9. Call peaks independently for every biological sample using paired-fragment mode. Check peak counts, peak widths, and FRiP alongside the earlier QC metrics when reviewing outliers.
10. Construct a declared consensus universe and count paired fragments in an explicitly recorded sample order.
11. Prove that count-matrix columns exactly match the metadata rows. Check that the proposed model matrix is full rank.
12. Choose and record the low-count rule before testing. Fit the model with the intended reference level and contrast; apply multiple-testing correction.
13. Inspect size factors, PCA, sample distances, dispersion behavior, and effect estimates. Treat these as diagnostics, not significance tests.
14. Annotate and enrich only with a suitable tested-region background. If the primary analysis has no FDR-significant regions, report that result first.
15. Preserve parameters, versions, logs, checksums, small summaries, figures, and limitations. Do not publish controlled data or credentials.

Proceed to the next stage only after the current stage's expected outputs and validation gate have passed.

---

Tutorial contents · Start the full tutorial: Overview →