mcDETECT Tutorial

MERSCOPE-focused walkthrough: mcDETECT setup, granule detection, profiling, manual subtyping, WT vs AD density, and Isocortex neuropil subdomains.

Tutorial

• mcDETECT tutorial: MERSCOPE analysis (step by step)
  • Analysis roadmap
  • Pages
• 1. What mcDETECT expects
  • 1.1 Platform type
  • 1.2 Transcript table (pandas.DataFrame)
• 2. Step A — Construct mcDETECT and understand every parameter
• 3. Step B — Granule detection (pattern from 3_detection.py)
  • Step B1 — Load inputs (conceptual)
  • Step B2 — Optional “rough” detection
  • Step B3 — “Fine” detection (recommended)
  • Step B4 — Optional region labels
• 4. Step C — Profile granules (profile)
• 5. Step D — Manual granule subtyping only (benchmark_subtyping.ipynb)
  • Step D1 — Normalize for clustering
  • Step D2 — Choose k and run MiniBatch k-means
  • Step D3 — Heatmap-driven biology
  • Step D4 — Manual mapping dictionary
  • Step D5 — Paired WT + AD objects (if applicable)
• 6. Step E — Granule density comparison WT vs AD (benchmark_subtyping.ipynb)
  • Step E1 — Spatial reference: spots per sample
  • Step E2 — Density definition (50 µm grid by default in the notebook helpers)
  • Step E3 — AD capture-efficiency correction
  • Step E4 — Per-spot counts for statistics
  • Step E5 — Export
• 7. Step F — Neuropil subdomain analysis, Isocortex (7_neuropil_subdomains.ipynb)
  • Step F1 — Region and grid constants
  • Step F2 — Inputs (prepared elsewhere in the pipeline)
  • Step F3 — Hard embedding (spot_embedding)
  • Step F4 — Drop empty spots
  • Step F5 — Hard normalized embedding for clustering
  • Step F6 — K-Means with k = 4
  • Step F7 — Visualization and follow-ups (optional)
• 8. Quick reference checklist