Gradient Biotech

Integrated cohort evidence

Use this workflow after multiple oncology analyses have completed and you need one ranked evidence table across immune, communication, mutation, CNA, mutational signature, ligand activity, and survival outputs. It is the use case for moving from separate pipeline results to a cross-modal translational evidence review.

Research question

Which completed analysis signals repeatedly support a responder, survival, mutation, copy-number, or treatment-arm hypothesis across the cohort?

Use case

Use this for responder versus non-responder review, treatment-arm comparison, mutation/CNA clinical enrichment review, signature group summaries, ligand activity hypothesis triage, and survival-linked feature screening. The workflow helps identify which findings recur across source runs while preserving source-run traceability.

Suggested path

  1. Run the relevant oncology analyses: immune profile, communication, mutation landscape, ligand activity, survival, and predictive biomarker modeling as appropriate.
  2. Run cohort_comparison for the study or select explicit source run IDs.
  3. Review the multi-omic driver evidence table first — it re-groups the gene- and pathway-shaped evidence rows (differential mutation, CNA enrichment, driver-pathway enrichment) by gene/pathway, summing each one's contributing evidence into a single ranked score, so a gene supported by mutation and copy-number and pathway evidence surfaces once instead of as separate rows. Expand a row to see its individual contributing evidence rows (domain, comparison, score, q-value).
  4. Review the full integrated evidence list below it for the complete cross-modal picture, including immune, communication, and survival rows the driver evidence table intentionally excludes — source pipeline names, comparison labels, effect values, p-values, adjusted q-values, and ranking scores.
  5. Drill back to source runs when a row needs detailed review.
  6. Use interpretation to summarize cross-modal evidence with metric citations.

Outputs to cite

  • evidence source pipeline
  • comparison label
  • feature, gene, or pathway name
  • effect size or score
  • p-value and adjusted q-value when available
  • integrated evidence rank (or combined driver-evidence score, for gene/pathway rows)
  • source run ID