Skip to content

11. Interpretation and reporting

← Previous: Annotation and enrichment · Tutorial contents · Next: Troubleshooting →

Report in this order

  1. Experimental design, samples, exclusions, and reference build.
  2. Processing and QC decisions with versions and parameters.
  3. Consensus definition, count unit, filter, model, reference level, and contrast.
  4. Number of regions entering and passing each gate.
  5. Primary multiple-testing-controlled result and effect-size threshold.
  6. Diagnostic figures and limitations.
  7. Clearly separated exploratory annotation or enrichment.

Language for common outcomes

Prefer “no region met the prespecified threshold” over “there was no biological difference.” A non-significant result can reflect limited power, heterogeneity, measurement noise, small effects, or a genuinely weak contrast.

Do not describe nearest genes as targets, motifs as active transcription factors, visual coverage differences as normalized group effects, or nominal p-values as discoveries.

Reproducibility package

Preserve metadata, input checksums, parameters, scripts, software versions, commands, validation summaries, compact result tables, figure sources, and a limitations document. Large or controlled data should remain in an appropriate archive rather than Git. Make public-data accessions and regeneration steps discoverable.

Copy-and-adapt provenance commands

Run these commands from the project root after activating the same environment used for analysis. The environment/local/ directory is ignored by this repository so machine-specific package URLs are not published accidentally.

set -euo pipefail

mkdir -p environment/local logs/reporting
conda list --explicit > environment/local/conda_explicit_spec.txt
conda env export > environment/local/conda_environment_resolved.yml
Rscript -e 'sessionInfo()' > environment/local/R_sessionInfo.txt

sha256sum \
  config/local/samples.tsv \
  config/local/input_manifest.tsv \
  config/local/analysis_parameters.yaml \
  results/consensus/consensus.bed \
  results/counts/consensus_counts.clean.tsv \
  results/differential_accessibility/all_tested_regions.tsv \
  > logs/reporting/key_output_sha256.txt

date '+%Y-%m-%dT%H:%M:%S%z' > logs/reporting/reporting_timestamp.txt

If a listed output does not exist because that optional stage was not run, remove only that path from the checksum command and state why it is absent. Keep the exact commands used for each sample and stage alongside these records.

The worked example's public reporting choices can be reviewed in docs/METHODS.md, docs/LIMITATIONS.md, and docs/REPRODUCIBILITY.md.

← Previous: Annotation and enrichment · Tutorial contents · Next: Troubleshooting →