Review Papers Against Their Own Results

 Long Explanation Copied

Verdict: This 2017 review (Hawi et al., American Journal of Medical Genetics Part B) convincingly assembles cross-disorder genetic evidence that cadherin genes — especially CDH13, CDH7, CDH11 and the 5q31–q35 protocadherin cluster — are recurrently implicated in ADHD, ASD, schizophrenia, bipolar disorder, and MDD, then adds genuine value by computationally prioritizing variants for functional follow-up Hawi et al. 2017 cadherin psychiatric genetics review.

Evidence assembled by the paper

Key associations reported (Table 1) are almost all subthreshold — no cadherin signal reaches genome-wide significance (5×10⁻⁸). The strongest are the CDH9/CDH10 5p14.1 locus in ASD (Wang 2009, p = 2.1×10⁻¹⁰) and CDH13 in ADHD (Lesch 2008, p = 7×10⁻⁶), the latter reinforced by a 16q23.1 linkage peak (p = 0.00034) and convergent mouse/cell work showing CDH13 modulates GABAergic interneuron activity Reported cadherin association p-values across five disorders. Enrichment analysis places cadherin genes top in cell–cell adhesion (GO:0007156, p = 2×10⁻⁹) and adherens-junction organization (GO:0034332, p = 3×10⁻⁵), consistent with PGC pathway findings where cadherins comprised 10 of 12 genes in the cell–cell junction pathway Gene enrichment results (Table 2).

Strengths and critique

Strengths: cross-disorder framing predating large PGC cross-disorder papers; a transparent in silico pipeline (42 SNPs → 1,610 with r²≥0.5 LD → 36 functional via CADD≥10/GWAVA≥0.5, 28/36 noncoding) with concrete mechanistic leads (rs79490173 disrupting miR-33 binding in the CDH7 3′UTR; CTCF/EBF1 motif disruptions at CDH13) and a falsifiable experimental roadmap Variant prioritization pipeline and functional predictions. Weaknesses: the review is inherently correlational — associations span heterogeneous populations, phenotypes and LD structures; ~70% of the underlying signals would likely not survive modern multi-million-sample GWAS; publication and winner's-curse bias are acknowledged but not quantified; CADD/GWAVA cutoffs are arbitrary and predictions unvalidated in this paper; CDH13 dominance may partly reflect study attention (citation bias). The paper does honestly flag that functional genomic validation is required. BGPT inference: treat the eight prioritized SNPs as hypotheses, not findings.

What would change this conclusion: null results in luciferase/EMSA/CRISPR-neuron assays for the eight shortlisted variants, or failure of CDH13 signals to replicate in later large biobank GWAS.

Explore further: CDH13 Best Evidence Graph the Associations

Author reviews: Ziarih Hawi Janette Tong Callum Dark Hannah Yates Beth Johnson Mark A Bellgrove

 BGPT Paper Review

Study Novelty

A 2017 synthesis of mostly pre-2015 association literature with a new in silico variant-prioritization layer; the cross-disorder cadherin framing and CADD/GWAVA triage were timely but not conceptually groundbreaking.

Scientific Quality

Transparent methods, structured tables, honest limitation discussion, and a clear validation roadmap raise quality; capped by reliance on subthreshold heterogeneous associations, arbitrary annotation thresholds, and absence of any experimental validation.

Study Generality

Spans five major psychiatric disorders and a 110+ gene superfamily, connecting adhesion biology to neurodevelopment broadly relevant to complex-trait genetics.

Study Usefulness

Useful as a gene-prioritization resource: eight specific SNPs with named follow-up assays give experimenters actionable targets, though clinical utility remains nil until functional validation.

Study Reproducibility

Pipeline tools (g:Profiler, CADD, GWAVA, RegulomeDB, HaploReg) are public and parameters stated (r²≥0.5, CADD≥10, GWAVA≥0.5), but the 42-SNP input list is in supplementary tables and thresholds/database versions were not version-locked.

Explanatory Depth

Offers plausible mechanistic hypotheses (TF motif disruption, miRNA sequestration, CDH13/GABAergic biology) but stops short of demonstrating any causal mechanism; depth is hypothesis-generating rather than mechanistic proof.