Complete publication workflow

1. Pre-specify the analysis

Before inspecting OmicsBraid outcomes, define the biological contrast, effect direction, ordered omic layers, inclusion/coverage rules, equivalence margin, trajectory margin, and multiplicity-control strategy.

2. Prepare analysis-ready assays

Use assay-appropriate normalization/QC outside OmicsBraid. Preserve sample identifiers and genuine missingness.

3. Construct and validate the object

obj <- omics_braid_data(
  assays = list(RNA = rna, Protein = protein, Phosphoprotein = phospho),
  metadata = metadata,
  sample_id = "sample_id"
)
validate_omics_braid_data(obj)

4. Run the primary analysis

fit <- run_omics_braid(
  obj,
  group = "group",
  reference = "WT",
  comparison = "Mutant",
  omic_order = c("RNA", "Protein", "Phosphoprotein"),
  bootstrap_B = 300,
  bootstrap_shrinkage = 0.05,
  ci_method = "analytic",
  integrated_ci_method = "analytic",
  equivalence_margin = 0.30,
  trajectory_margin = 0.15,
  pattern_draws = 1000,
  seed = 1
)

5. Build auditable tables

master <- braid_results_table(fit)
write_omics_braid(fit, "omicsbraid_output", save_plots = TRUE, top_n = 30)

6. Use empirical calibration appropriately

For targeted robustness checks, use subject-level permutation omnibus and null-shift bootstrap heterogeneity. Do not describe post-screening empirical results as independent genome-wide confirmation.

7. Report multiple inferential layers

For key entities report layer-specific effects with intervals, GLS consensus, omnibus evidence, Q_omics, braid pattern/status, and any empirical sensitivity result. Avoid reporting a braid label without the underlying effects.