Review Papers Claim by Claim

Visual review & critique — "Genetic drivers of m6A methylation in human brain, lung, heart and muscle" (Xiong et al., Nat Genet 2021)

Critical appraisal (visual first, rationale second)

Detailed critique and recommendations (concise)

1. QC and normalization: PEER factor selection and IP/GC corrections are appropriate; authors demonstrated inflection points for PEER removal to maximize g‑m6A detection (methodologically sound). Cite: paper methods and Fig.2a,b Xiong et al., normalization/PEER
2. QTL mapping: FastQTL permutation with empirical P‑values is standard; authors used locus‑wide correction and stricter 2‑step correction for brain high‑confidence calls — appropriate given multiple-testing differences across tissues.
3. Tissue specificity claim: Robust: directionality-consistency analyses and simulations argue observed specificity isn't purely power-driven — but sample-size differences (brain>>heart) still affect absolute sharing estimates; single-cell and ancestry‑diverse cohorts would reduce blindspots.
4. GWAS interpretation: Valuable: many colocalizations place noncoding GWAS SNPs into an RNA‑level mechanism; but each locus needs locus‑by‑locus functional follow‑up (allele-specific reporter assays, CRISPR base editing, or engineered isogenic lines measuring m6A and expression). The paper provides hypotheses; not definitive causal demonstration.
5. Trans‑regulator predictions: Predicting RBPs from CLIP enrichments + PPI + expression correlations is sensible; these are priority candidates but require perturbation (knockdown/overexpression + m6A mapping) and biochemical validation to move from candidate to regulator.

Where the data could mislead or be incomplete

• Antibody cross-reactivity and IP efficiency variability can create false peaks or mask true sites; single‑nucleotide resolution methods (GLORI, miCLIP alternatives, or direct‑RNA nanopore detection) would validate key loci Xiong et al., meRIP limitations
• Bulk tissue hides cell-type–specific QTLs; a brain‑specific m6A QTL might act only in excitatory neurons whereas bulk signal shows tissue specificity — single‑cell or nuclei‑based meRIP would be transformative.
• Ancestry differences: discovery dataset mainly EUR (GTEx), validation was Asian LCLs; cross‑ancestry generalizability needs explicit testing. The LCL comparison showed very low sharing, consistent with cell‑type differences but also ancestry/LCL artefacts.

Actionable recommendations for follow-up studies

1. Targeted validation: pick ~10 high‑confidence colocalized GWAS–m6A QTLs and perform allele‑specific reporter assays, nanopore direct RNA m6A calling, and CRISPR base editing in relevant primary cell models.
2. Trans‑regulator tests: perturb top predicted RBPs (e.g., UPF1, NCBP2, LARP4, ATXN2) in relevant tissues/cell types, measure aggregated m6A target levels and single‑site m6A changes; test physical interactions with writers/erasers.
3. Single‑cell extension: develop low‑input antibody‑free or improved single‑cell m6A assays to resolve cell‑type QTLs and compare tissue vs cell‑type specificity.
4. Population diversity: expand to non‑EUR cohorts to identify ancestry‑specific g‑m6A loci and test generality of GWAS colocalizations.

Summary judgement (short)

This paper delivers a major resource — tissue‑resolved m6A maps and the first catalog of genetically driven m6A QTLs in key human tissues; methods are sound and results well‑documented, but mechanistic claims remain largely hypotheses that require locus‑specific functional validation and single‑cell/cell‑type resolution to remove bulk/tissue confounders Xiong et al., resource value and caveats

Author reviews (quick links)