# C8 spatial projection record (06) — 2026-10-30
## Data and methods
- GSE222987: 16 Visium samples (NC1a villous control/NC1b decidua/NC1c fetal membranes/NC2-4
  controls/9 COVID-19/2 IUFD), 18,578 spots
- Official tissue_positions/scalefactors not provided → grid coordinates decoded from Visium barcode
  geometry (x 0-63, y 0-63, col=2x+y%2 hexagonal arrangement, empirically validated)
- h5 var_names carry the GRCh38______ prefix, stripped (root cause of empty signature columns on
  first run)
- Signatures: S1 concordant (50), S2 reversed (50), S3 macrosomia-specific up/down (50 each), from
  03_decoupling_table (C7 v2)

## Results
1. Cross-spot signature–cell-type correlations (mechanistic localisation):
   - S1 immune activation: negatively correlated with SCT (-0.46), positively with fibroblast (0.26)
     → localised to stromal/immune niches rather than trophoblast
   - S2 reversed signature: CTB 0.20 / EVT 0.20 / Fib 0.37, SCT -0.36 → trophoblast progenitor–
     stromal niche
   - S3 macrosomia up: **Endothelial 0.37, CTB 0.24** → vascular/growth programme compartment
2. Sample level: S2 high in decidua/fetal-membrane samples (tissue-compartment effect, not disease);
   within villous controls, S3_down higher in NC1a/NC2 — suggesting the macrosomia down-signature
   may partly reflect a conserved programme of healthy villi; interpret with caution
## QC judgement
- Spatial projection is methodologically feasible (barcode decoding + score_genes); no GDM control
  spatial data exist, so it can serve as "localisation" but not "comparison"
- Figures: 06_spatial_NC1a_signatures.png, 06_spatial_sample_heatmap.png
