Review Claim by Claim

Key positive findings (what the paper does well)

• Comprehensive synthesis of modern V2F toolset: multi‑ancestry fine‑mapping, single‑cell chromatin/transcriptome maps, 3D contact assays and high‑throughput perturbations (MPRAs, CRISPRi/a, base editing) — presented clearly and with up‑to‑date references Maynard et al. 2025 - methods summary
• Realistic appraisal of limits: LD, context dependence (cell state, developmental stage, stimulation), and low eQTL colocalisation rates (<50%) are acknowledged and explained Colocalisation & context
• Concrete locus examples (TCF7L2, MTNR1B, SLC16A11, PNPLA3, FTO) link genetic evidence to plausible molecular mechanisms across tissues — strengthens translational relevance Locus vignettes

Constructive criticisms / blindspots

• Review is descriptive: the field now needs standardised quantitative benchmarks (e.g., systematic comparison of fine‑mapping+annotation pipelines against held‑out validated variants) — the paper advocates this but does not present new benchmarking data Call for benchmarks
• Relative paucity of non-European examples in functional validation discussion: while the review cites multi‑ancestry GWAS, many functional assays remain performed primarily in cells/lines of European-ancestry donors or generic cell lines — the authors flag this but greater emphasis on ancestry-specific functional pipelines and datasets would strengthen generalizability Need for diversity in V2F
• Under-addressed negative results / publication bias: the field has an unknown rate of failed validation attempts; the review mentions publication bias but cannot quantify it — an explicit recommendation: publish negative CRISPR/MPRA datasets to improve priors and training of sequence models.
• Translational claims should be guarded: while loci like SLC16A11 and PNPLA3 present converging evidence, most GWAS loci remain without therapeutic-grade effector gene assignments — the review rightly resists over‑claiming, but readers must not assume broad near-term therapeutic impact.

Concrete recommendations (what to do next)

1. Construct community benchmarking sets: a curated list of ~100 loci with orthogonal validation (orthogonal = coding evidence, base editing replication, in vivo phenotyping) to evaluate combined statistical+functional pipelines.
2. Systematically expand single‑cell eQTL/caQTL panels across ancestries and stimulus states (lipid, glucotoxicity, cytokines) — prioritize islets, hepatocytes, adipocytes, muscle and vascular cell types.
3. Adopt pre-registered negative-result repositories for MPRA/CRISPR screens (format: guide/sequence, cell type, effect size distribution) to reduce publication bias and improve ML training.
4. Invest in cross-tissue organoid co-cultures and organ-on-a-chip experimental pipelines to measure cross‑tissue integrative effects highlighted by adipose/liver cross-talk (e.g., PNPLA3, PPARG loci).

Evidence anchors — selected citations used in this analysis

• Maynard et al. 2025 (review)
• Suzuki et al. 2024 (GWAS)
• Gaulton et al. 2015 (fine-mapping)