# C16 spatial projection v2 (control villi only + within-sample correlations)

## 1. Sample scope
- Included: 4 healthy controls (NC1a_villi, NC2_control, NC3_control, NC4_control)
- Excluded: 12 samples (8 COVID + 2 IUFD + decidua + fetal membranes)
- The original version included all 16 samples and pooled correlations across all spots
  → COVID/IUFD/compartment confounding

## 2. Signature definitions (v2 decoupling table)
- S1_concordant_up: n=50
- S2_reversed_upInHyp: n=50
- S3_macrosomia_up: n=50

- GSM6937756 NC1a_villi: 548 spots → 529 villous spots retained (97%)
- GSM6937769 NC2_control: 2,027 spots → 2,020 villous spots retained (100%)
- GSM6937770 NC3_control: 2,605 spots → 2,604 villous spots retained (100%)
- GSM6937771 NC4_control: 1,894 spots → 1,888 villous spots retained (100%)

## 3. Signature × cell-type correlations (computed per sample, pooled by Fisher z)

| Signature | Cell type | Per-sample r | Pooled r (95%CI) | Heterogeneity I² |
|---|---|---|---|---|
| S1_concordant_up | SCT | -0.13, +0.01, -0.37, -0.38 | -0.25 (-0.27, -0.23) | 99% |
| S1_concordant_up | CTB | -0.10, +0.00, +0.12, -0.15 | -0.00 (-0.02, +0.02) | 96% |
| S1_concordant_up | EVT | -0.01, -0.06, -0.03, -0.32 | -0.12 (-0.14, -0.09) | 97% |
| S1_concordant_up | Hofbauer | +0.13, +0.03, +0.27, +0.31 | +0.20 (+0.18, +0.22) | 97% |
| S1_concordant_up | Endothelial | -0.10, -0.05, -0.06, +0.06 | -0.03 (-0.05, -0.01) | 87% |
| S1_concordant_up | Fibroblast | +0.21, -0.01, +0.18, +0.36 | +0.18 (+0.15, +0.20) | 98% |
| S2_reversed_upInHyp | SCT | -0.34, -0.24, -0.42, -0.54 | -0.40 (-0.42, -0.38) | 98% |
| S2_reversed_upInHyp | CTB | -0.20, -0.04, +0.31, -0.25 | +0.02 (+0.00, +0.05) | 99% |
| S2_reversed_upInHyp | EVT | -0.18, -0.05, +0.20, -0.43 | -0.07 (-0.10, -0.05) | 99% |
| S2_reversed_upInHyp | Hofbauer | +0.28, +0.12, +0.17, +0.37 | +0.22 (+0.20, +0.24) | 97% |
| S2_reversed_upInHyp | Endothelial | -0.22, +0.16, -0.14, +0.11 | +0.01 (-0.02, +0.03) | 98% |
| S2_reversed_upInHyp | Fibroblast | +0.65, +0.34, +0.11, +0.69 | +0.40 (+0.38, +0.42) | 100% |
| S3_macrosomia_up | SCT | +0.30, +0.00, +0.22, +0.22 | +0.16 (+0.14, +0.19) | 96% |
| S3_macrosomia_up | CTB | +0.54, +0.43, +0.30, +0.51 | +0.42 (+0.40, +0.44) | 96% |
| S3_macrosomia_up | EVT | +0.10, +0.11, +0.01, +0.24 | +0.11 (+0.09, +0.13) | 95% |
| S3_macrosomia_up | Hofbauer | -0.30, -0.11, -0.17, -0.24 | -0.18 (-0.20, -0.16) | 89% |
| S3_macrosomia_up | Endothelial | +0.08, +0.12, +0.11, +0.09 | +0.10 (+0.08, +0.13) | 0% |
| S3_macrosomia_up | Fibroblast | -0.58, -0.16, -0.29, -0.46 | -0.33 (-0.35, -0.31) | 98% |

## 4. Interpretation rules

- The above is expression localisation in **healthy/control placenta**, not disease localisation;
  GSE222987 contains no GDM cases.
- Per-sample computation avoids cross-spot pseudoreplication; high I² indicates large between-sample
  heterogeneity and the pooled values are indicative only.
- Spatial autocorrelation was not modeled, so **no P values are reported** — effect sizes and
  confidence intervals only.
- The signatures derive from bulk DE, so their correlation with cell-type scores is partly
  **circular** (signature genes are themselves cell-type markers); this is a coarse
  expression-compartment attribution, not independent mechanistic evidence.

## 5. Key differences from the original version (overturning the original conclusion)
- Original: "S3 macrosomia signature localises to endothelium (+0.37)" (all 16 samples concatenated,
  including 8 COVID + 2 IUFD)
- Corrected: S3 ↔ Endothelial **r=+0.10 (95%CI 0.08–0.13, I²=0%)**, essentially zero
  → S3 actually localises to **CTB (r=+0.42, consistent +0.30 to +0.54 in 4/4 samples)**
  → the original "endothelial localisation" was an artifact created by the COVID/IUFD samples and
  cross-spot pooling, and **must be deleted**
- S2 (reversed/ECM) ↔ Fibroblast r=+0.40 (positive in 4/4 samples, but I²=100%, extreme
  heterogeneity)
  → consistent with the ECM enrichment and with the fibroblast-composition confounding concern
  (P0-10); a compositional explanation cannot be excluded
- S1 (concordant) ↔ SCT r=-0.25, Hofbauer r=+0.20 → the immune/IFN signature leans towards
  Hofbauer, but the signal is weak

## 6. Wording to use in the main text
- "Projection onto four control placental villous samples" (not healthy atlas / 16 samples)
- "Signature scores correlated with cytotrophoblast markers (S3: r=0.42) and fibroblast markers
  (S2: r=0.40), with substantial between-sample heterogeneity (I²≈100%)"
- State explicitly: "This is an expression-compartment mapping in control tissue, not disease
  localization, and is partly circular because bulk-derived signatures include cell-type marker
  genes."
