Multi-omic integration

Connect single-cell, bulk, mutation, repertoire, and clinical data in one oncology study — from molecular profiling through survival analysis and analysis.

Research question

How do TME composition, immune phenotype, mutational features, and signaling patterns combine to explain treatment response and clinical outcomes across my cohort?

Who this is for

Computational oncology groups supporting multi-modal tumor studies
Translational teams running correlative analyses on clinical trial biopsies
Pharma programs integrating sequencing, expression, and outcome endpoints

Data requirements

Modality	Source	Purpose
Single-cell or bulk RNA-seq	Compbio + oncology	TME profiling, immune deconvolution, communication
MAF / VCF	Oncology	Mutation landscape, TMB, signatures
TCR/BCR repertoire	Oncology	Clonotype and diversity analysis
Clinical endpoints	Oncology	Survival, response, recurrence
Spatial transcriptomics	Compbio	Spatial TME mapping
Whole-slide images	Pathology	Tissue morphology context

The key integration layer is aligned patient/sample keys across all modalities within one study.

Workflow

Create study → Register samples, endpoints, and datasets
  → TME profiling [compbio]
  → Immune profile + cell communication [oncology]
  → Mutation landscape [oncology]
  → Survival with molecular features [oncology]
  → Multi-modal AI interpretation

Step 1 — Organize the cohort

Create an oncology study with tumor type, stage, and treatment arm metadata. Register all samples with consistent patient_id and external_id keys. Add clinical endpoints for every patient with follow-up data.

See Data upload.

Step 2 — Run molecular pipelines

Execute pipelines appropriate to each modality:

Modality	Pipeline	Page
Single-cell	TME composition (compbio-backed)	TME
Bulk RNA-seq	Immune profile	Immune
Single-cell	Cell-cell communication	Communication
MAF	Mutation landscape	Mutations

Run pipelines in any order; each produces independent artifacts linked to the study run history.

Step 3 — Integrate features for survival

Combine outputs into a feature matrix — TMB, immune phenotype, exhaustion score, CD8 fraction, dominant signature — keyed by patient_id. Launch survival analysis with feature_path and cox_covariates:

curl -X POST http://localhost:8001/oncology/jobs/survival \
  -H "Content-Type: application/json" \
  -d '{
    "study_id": "your-study-id",
    "parameters": {
      "clinical_path": "/path/to/clinical.csv",
      "feature_path": "/path/to/multi_omic_features.csv",
      "stratify_by": "treatment_arm",
      "cox_covariates": ["tmb", "cd8_score", "exhaustion_score"]
    }
  }'

For longitudinal studies, include a timepoint trajectory file to track biomarker changes across treatment course.

Step 4 — Multi-modal interpretation

Launch interpretation with all completed run IDs:

curl -X POST http://localhost:8001/oncology/jobs/interpret \
  -H "Content-Type: application/json" \
  -d '{
    "study_id": "your-study-id",
    "parameters": {
      "source_run_ids": [
        "onco-run-communication",
        "onco-run-immune",
        "onco-run-mutation",
        "onco-run-survival"
      ],
      "interpretation_type": "summary"
    }
  }'

The interpret pipeline synthesizes findings across communication, immune, mutation, and survival outputs with metric citations.

Expected outputs

Unified run history with provenance for every pipeline layer
TME composition, signaling network, immune phenotype, and mutation landscape artifacts
Survival curves and Cox models with multi-omic covariates
Integrated AI narrative suitable for translational reports and collaborator review
Exportable oncoprint, KM curve, and signaling graph data for publications

Example end-to-end scenario

A phase II checkpoint inhibitor trial with pre-treatment biopsies:

scRNA-seq → TME shows elevated regulatory T cells at the tumor-stroma interface
Communication → LGALS9-HAVCR2 is the dominant suppressive axis in non-responders
Immune profile (bulk) → Non-responders classified as immune-excluded with high terminal exhaustion
Mutation landscape → Low TMB and APOBEC-dominant signature in non-responders
Survival → TMB and CD8 score independently predict PFS in multivariate Cox model
Interpret → Grounded summary connecting immune exclusion, suppressive signaling, and low neoantigen burden to IO resistance

Reproducibility

Every step records pipeline version, parameters, and artifact paths. Re-run any layer independently without losing provenance for the full analysis chain.

See Runs and reproducibility.

Menu

Documentation