Review Papers Claim by Claim

A. Study snapshot (visual)

B. Key quantitative claims and immediate interpretation

1. "No CpG survived FDR" — the authors report that after multiple-testing correction no sites remained significant; they therefore used an effect-size + nominal p-value filter (effect > 0.2 and p < 0.05) to prioritize loci; this raises the probability of type I errors and means findings are exploratory (Multiple-testing and thresholds
2. "24 maternal / 18 infant CpGs, 1 shared" — suggests largely distinct maternal and fetal DNAm responses to the same exposure; plausible biologically, but low sample size limits generalization (Result counts
3. "TH as shared gene" — authors report TH (tyrosine hydroxylase) nearest to the shared CpG and note prior animal data showing PM2.5 effects on TH expression; this is hypothesis-generating for neuroendocrine links but requires replication and functional validation (TH annotation

C. Strengths (what the paper does well)

• Clear, transparent methods: array type, QC steps, exposure window choice, and that FDR produced no hits are explicitly reported (Transparent methods
• Paired mother–infant dyad design enables direct comparison across maternal–fetal interface (rare in many exposure–epigenetic studies) (Dyad study design
• Use of a targeted high-value allergy/asthma CpG array increases prior biological plausibility for immune-related hits (but trades off genome-wide discovery) (Targeted CpG panel rationale

D. Major limitations & biases (critical)

• Severe underpowering and multiplicity: n=10 dyads → low power to detect small DNAm differences and impossible to correct for multiple testing at genome-level significance; the authors acknowledge this (Sample-size limitation
• False-positive risk from post-hoc thresholding: using effect-size + nominal p filters after FDR failure inflates false discovery probability; such lists should be labeled strictly exploratory and validated in independent cohorts (Thresholding caution
• Confounding and exposure misclassification risk: exposure estimated from residential proximity to monitors (inverse-distance-squared) — standard but different from personal exposure monitors; unmeasured confounders (SES, time-activity, indoor sources) can bias associations (Exposure assignment limits
• Tissue and cell-type complexity: PBMCs/CBMCs are mixed-cell populations; although PCA of covariates was performed, no cell-deconvolution or cell-type–specific analyses are reported — cell-composition differences can drive apparent DNAm associations (Cell-type confounding potential
• Generalisability: small, regionally specific (Fresno/Clovis) and demographically skewed (7/10 Hispanic) sample limits external validity.

E. Reproducibility checklist

1. Methods described: DNA assay, QC pipeline (minfi), exposure assignment, and regression approach — present in Methods (Methods transparency
2. Data availability: data available upon reasonable request (not public); for reproducibility, full processed matrices and code should be shared in a repository (missing) (Data access statement
3. Missing for immediate reproduction: (a) per-sample PM2.5 values and IDs, (b) processed beta/M-value matrices, (c) R scripts or notebooks for regression and filtering steps; authors could release these to improve reproducibility.

F. How convincing is the biological story?

The claim that maternal and fetal circulations show largely distinct DNAm responses to the same PM2.5 exposure is plausible (different cell compositions, placental filtering, maternal physiology) and consistent with other work showing tissue-specific epigenetic responses to air pollution (placenta, cord blood, PBMCs) — but the present evidence is preliminary and insufficient to support mechanistic claims without replication and functional data (Placental methylation example

G. Concrete recommendations to authors / for follow-up studies

1. Replication in a larger, independent cohort (n > 100 dyads) using the same targeted array and pre-registered analysis plan to test the 42 candidate CpGs — pre-registration avoids HARKing and reduces false-positive risk.
2. Make processed DNAm matrices, per-sample PM2.5 values, code, and covariate tables publicly available (e.g., GEO/ArrayExpress + GitHub) to permit independent re-analysis.
3. Conduct cell-deconvolution (Houseman or cord-blood–specific methods) or sort cell types (CD4, CD8, monocytes) to test whether associations are cell-composition driven.
4. Perform technical validation of prioritized CpGs (pyrosequencing or targeted bisulfite sequencing) in the same samples and then in replication cohort(s).
5. Integrate expression data (RNA) or chromatin accessibility (ATAC) for prioritized loci (e.g., TH) to link methylation changes to function.
6. Use personal exposure monitors and/or include PM2.5 composition (black carbon, organics) to refine exposure-response relationships and reduce exposure misclassification.

H. What would disprove the paper's main claims?

• Failure to replicate the maternal and infant CpG associations in an adequately powered independent cohort using identical QC and analysis (i.e., none of the 24/18 CpGs or the shared CpG replicate).
• Demonstration that the nominal associations are fully explained by cell-composition differences, unmeasured confounders (e.g., household smoke, occupation), or technical batch effects after appropriate adjustment.

I. Quick linkage to high-relevance supporting literature

• Context on PM2.5 and epigenetics in placenta: see placental global methylation study (Janssen et al., 2013 (placenta)
• Single-cell chromatin/cytokine pregnancy work showing pregnancy modifies pollutant–epigenome relationships in immune cells (supports tissue- and pregnancy-specific effects) (Prunicki et al., 2024 (single-cell epigenetics)

J. Short actionable checklist for readers evaluating similar small epigenetic studies

1. Check whether effect estimates are reported with SE/CIs and whether FDR or other robust multiple-testing corrections were attempted.
2. Look for per-sample exposure distributions and processed data availability to judge exposure variability and potential leverage points.
3. Confirm whether cell-composition adjustments or sorted-cell analyses were performed.
4. Favor findings replicated by independent cohorts or validated with orthogonal assays (pyrosequencing, targeted bisulfite sequencing).