Differential analysis¶

RNA differential expression runs at step 12. ATAC/ChIP differential chromatin analysis runs at step 14. Both use metadata-driven DESeq2 designs.

Automatic design¶

omnomnomics rna \
  -i EXPERIMENT \
  -g GRCh38 \
  -j 12 \
  -m metadata.tsv \
  --de-columns genotype,stim \
  --de-block donor \
  --de-interactions

Use --de-formula when the model must be specified directly. It overrides --de-columns and --de-block.

YAML configuration¶

Use a YAML file when an analysis needs more control than the design options on the command line. A compact genotype-by-stimulation configuration can look like this:

version: 1
io:
  out_dir: "genotype_by_stim"

design:
  formula: "~ donor + genotype * stim"
  reference_levels:
    genotype: "NT"
    stim: "C"

contrasts:
  mode: "explicit"
  explicit:
    items:
      - [stim, L, C]

filtering:
  enabled: true
  method: "min_count_samples"
  min_count: 10
  min_samples: 2

deseq2:
  lfc_shrink:
    enabled: true
    type: "apeglm"
    use_for_tables: true

thresholds:
  alpha: 0.05
  lfc_for_sig: 1.0

enrichment:
  enabled: true

design.formula defines the DESeq2 model, while reference_levels fixes the baseline used to interpret coefficients. Explicit contrasts name the comparison to report. In this interaction model, [stim, L, C] is the stimulation effect at the reference genotype (NT); genotype-specific response differences require an additional interaction-coefficient contrast. Filtering removes weakly detected features before fitting; thresholds controls significance classification and plotting rather than the model formula. Each analysis writes to its own io.out_dir.

The complete example configuration also documents automatic contrasts, QC and plotting controls, latent factors, clusterProfiler and decoupler enrichment, custom gene sets, and runtime policy. Values omitted from a user configuration retain the packaged defaults.

omnomnomics rna \
  -i EXPERIMENT \
  -g GRCh38 \
  -j 12 \
  -m metadata.tsv \
  --de-config de_grouped.yaml

Repeat --de-config to run several analyses sequentially. Each file must set a unique io.out_dir; do not combine repeated config files with global --de-out-dir.

The workflow writes resolved metadata and rendered R scripts so the statistical analysis can be audited and customized later.

Chromatin interpretation¶

After differential chromatin testing, omnomnomics creates BED sets for significant, increased, and decreased regions. It can then add signal profiles, heatmaps, and motif reports for those sets.

--post-de-signal-policy	Behavior
auto	create signal plots when the required BigWigs are available
require	fail if the requested signal plots cannot be produced
skip	omit signal plotting while retaining the differential results

Every signal and motif attempt is recorded in status tables. A motif status of TIMEOUT or FAIL therefore identifies an optional report that did not finish; the differential result tables and region sets remain available.