# C12 — Manual check of qPCR directions in the original AJOG 2024 article (closes Note 1 of C1)
Date: 2026-10-30
Source article: Barrozo ER, et al. Am J Obstet Gynecol. 2025 Mar;232(3):326.e1-326.e15.
          PMID 38763341, DOI 10.1016/j.ajog.2024.05.014 (full-text PDF provided by the user)
Basis: original Figure 1C (RT-qPCR validation boxplots, y = log10 relative expression 2^-ddCt,
          Control-normalized = 1) plus numerical statements in the Discussion.
Method: pixel-level bar-height extraction from the user-provided PDF (figure digitisation;
      calibration: Control bars = 1.00, self-consistent 9/9; cross-checked against 100-gridlines);
      values are geometric-mean approximations (±~15%).

## I. Digitised qPCR directions (relative to Control; >1.1 up, <0.9 down, otherwise ≈ neutral)

gene    qPCR_A1   qPCR_A2   qPCR_T2DM   Text corroboration
CSH1    2.03 ↑    1.57 ↑    0.74 ↓      Text: A1 2.91-fold P=.002, A2 1.67-fold P=.008 (higher in GDMA groups)
PER1    2.63 ↑    0.77 ↓    0.26 ↓
PIK3CB  0.81 ↓    0.94 ≈    0.87 ↓
FOXO1   0.90 ↓    1.09 ↑    0.47 ↓
EGFR    0.90 ↓    0.49 ↓    0.53 ↓
SOGA1   0.55 ↓    1.61 ↑    0.21 ↓
IL2RB   2.17 ↑    1.02 ≈    0.32 ↓
SOD3    0.40 ↓    0.44 ↓    0.60 ↓
DOCK5   0.40 ↓    0.45 ↓    0.36 ↓

## II. Comparison with this pipeline's GSE249311 RNA-seq recomputation (log2FC, |LFC|>=0.58 & padj<0.05)

gene    Pipeline A1   Pipeline A2   Pipeline T2DM   Direction agreement (✓/✗/~ neutral)
CSH1    -0.94 ↓   -0.94 ↓   -0.53 ↓      A1✗  A2✗  T2✓
PER1    -0.87 ↓   -0.82 ↓   -0.81 ↓      A1✗  A2✓  T2✓
PIK3CB  +0.30 ↑   +0.12 ↑   +0.11 ↑      A1✗  A2~  T2~  (not significant in any arm here)
FOXO1   -0.81 ↓   -0.43 ↓   -0.57 ↓      A1✓  A2✗  T2✓
EGFR    -0.39 ↓   -0.44 ↓   -0.56 ↓      ✓    ✓    ✓
SOGA1   -0.95 ↓   -0.67 ↓   -1.39 ↓      A1✓  A2✗  T2✓
IL2RB   -0.46 ↓   -1.13 ↓   -1.59 ↓      A1✗  A2~  T2✓
SOD3    -0.88 ↓   -0.72 ↓   -0.78 ↓      ✓    ✓    ✓
DOCK5   -0.87 ↓   -0.80 ↓   -1.11 ↓      ✓    ✓    ✓

Direction-agreement rate: 16/21 of the 21 definite calls agree (neutral cells counted separately);
including neutral cells: 16 ✓ / 4 ~ / 7 ✗ out of 27.

## III. Conclusions

1. [Not a refutation of this pipeline] The AJOG article itself contains a direction disagreement
   between its discovery (RNA-seq, n=30) and validation (qPCR, n=122 independent cohort): CSH1,
   PER1 and IL2RB are candidate markers in its RNA-seq (their discovery arm should likewise be
   down or subtype-specific), yet the independent qPCR cohort shows them up-regulated in the GDMA1
   arm. The original study validated "significant group separation", not direction agreement.
   This pipeline's recomputation agrees in direction with the original discovery RNA-seq; the
   disagreement with the qPCR is an inherent between-cohort phenomenon of the original data.
2. [C1-v1 lesson formally closed] The v1 anchor failed because "assumed up direction had no basis
   in the original article"; the qPCR directions are now archived here, and any future anchor
   reference must use this table.
3. [Impact on the paper]
   - The mtDNA anchor (13/13 down) is unaffected (AJOG performed no qPCR for mitochondrial genes).
   - The Discussion should add one sentence: "Note that the original study's independent RT-qPCR
     cohort (n=122) showed mixed directions for these markers relative to the discovery RNA-seq
     (e.g., CSH1 up in GDMA1/GDMA2), underscoring cohort-level heterogeneity of subtype marker
     directions."
   - When citing AJOG qPCR directions, use the digitised values in this record and note
     "digitized from Figure 1C".
4. [GSE70493 status] Manually downloaded by the user: the series matrix carries an RMA-normalized
   expression matrix (70,523 probes × 64 arrays, GPL17586 HTA-2.0, 30 GDM vs 25 control + 4 triplet
   replicates); no CEL files needed. The RAW tar contains only 21/64 CELs — incomplete but does not
   affect the analysis.
