Goal | Section A -- Cutoff sensitivity | A1. Trajectory classification cutoff (rho_cutoff) | A2. Three-axis weight sensitivity | Section B -- Tool comparison | Section C -- Negative control depletion | C1. Top-N depletion (negative controls absent from the most-prioritised candidates) | C2. Bottom-N enrichment (negative controls concentrated at the lowest-priority end) | Section D -- Held-out PDAC cohort projection (v0.99.12) | Section E -- Three-way baseline: where does enrichment come from? | Limitations of these analyses

Goal | 1. Why a hand-engineered rule (and not a supervised model) | 2. The 6-function API at a glance | 3. LTBP1 — the supported_uncertain exemplar | 4. GAPDH — the housekeeping negative control | 5. Anchor enrichment evaluation (post-freeze) | 6. Reading the case_study() output | 7. What the audit framework does not claim | Session info

Goal | Step 1 -- A simulated multi-cohort, multi-omic benchmark | Step 2 -- Depth-adjusted blood observability | Step 3 -- Rank candidates | Step 4 -- Does the ranking survive dropping a cohort? | Step 5 -- Comparators, scored honestly | Step 6 -- The frozen one-shot audit | Why a calibration check matters | Scope

Goal | Layer 1 — Offline, bundled | Layer 2 — User cohort | Layer 3 — Re-derive the bundled atlas | Layer 4 — From FASTQ | Provenance summary | Session information

Goal | Side-by-side audit summary | Detailed audit features per gene | Case 1 — LGALS3BP (high_confidence anchor) | Case 2 — LTBP1 (supported_uncertain, Class B inverse exemplar) | Case 3 — ALB (penalized, plasma high-abundance gate) | Case 4 — GAPDH (excluded, housekeeping gate) | Take-away | Session information

Goal | Step 1 — Load the bundled toy cohort | Step 2 — Fit the stage-aware RNA model | Step 3 — Fit the protein limma model | Step 4 — Assemble the user evidence table | Step 5 — Run the audit framework on user evidence | Step 6 — Bioconductor-native input via SummarizedExperiment (v0.99.2) | Step 7 — Single-patient trajectory alignment (v0.4.1) | Scope boundary | Appendix A — Stage harmonization (TNM / AJCC -> Normal / Early / Mid / Late) | Appendix B -- Serum projection (binary group contrast, no stage)

Goal | Step 1 -- Build a synthetic non-PDAC cohort | Step 2 -- Project through the framework (one call) | Step 3 -- What changes between PDAC and non-PDAC? | Step 4 -- Translation discipline transfers, but the join target changes | Limitations

Goal | 1. Single-gene query — query_gene() | 2. Multi-gene comparison — query_panel() | 3. Multi-parameter candidate listing — list_candidates() | 4. Filter audit trail — trace_filters() | 5. Plain-text summary — summarize_gene_evidence() | Atlas metadata | Next | Session