Crunching the numbers. Responsibly.
Most genetic variants associated with complex heritability phenotypes lie in non-coding regions and are thought to influence disease risk by regulating gene expression. However, most transcriptome-wide association approaches primarily model local (cis) genetic effects, leaving much of gene regulation unexplained. Here, we show that incorporating distal (trans) regulatory effects improves the prediction of gene expression and the identification of disease-associated genes. Using RNA sequencing data from six human post-mortem brain regions, we developed INGENE and MODULE, two models capturing the combined influence of candidate trans-acting variants within gene coexpression networks. Integrating these models with conventional cis-based predictors improved gene expression imputation (maximum likelihood estimation, α = 0.05) for 18,744 genes across regions. Applying this framework to Psychiatric Genomics Consortium wave 3 genotypes identified 766 genes associated with schizophrenia (PFDR < 0.01), including 641 not previously reported by transcriptome-wide analyses. These findings highlight the contribution of distal regulatory mechanisms and gene network interactions to schizophrenia risk.
Applying this framework to Psychiatric Genomics Consortium wave 3 genotypes identified 766 genes associated with schizophrenia
association design lacks independent schizophrenia replication and inherits evaluation-dependent gene selection
Integrating these models with conventional cis-based predictors improved gene expression imputation for 18,744 genes across regions
the pooled count uses evaluation-dependent filtering and exceeds the per-region cis–trans enhancement counts
incorporating distal (trans) regulatory effects improves the prediction of gene expression and the identification of disease-associated genes
coverage increases, but trans models have lower external per-gene accuracy and disease associations remain unvalidated
Derived from the full evaluation — not a separate score.
Strengths
INGENE and MODULE provide distinct coexpression-informed routes to trans prediction and are integrated with CIS and EpiXcan. The study also compares its approach with BGW-TWAS and MOSTWAS in DLPFC.
↳ Introduction; Results, “Evaluation of cis and trans model training performance”
Models are examined across LIBD, GTEx, and CMC brain datasets, and the combined predictor is evaluated in CMC after fitting in GTEx. The results transparently show that cis models retain higher per-gene external adjusted R².
↳ Results, Figures 3 and 4c
The integrated framework is applied across 62 PGC3 cohorts comprising 102,613 individuals, with conditional analysis and leave-site-out consistency assessment. This creates a substantial hypothesis-generation resource for psychiatric genetics.
↳ Results, “coTWAS identifies trait associations in PGC3 cohorts”; Table 2; Supplementary Figure 18
Limitations
Trans genes are retained using agreement with models trained in GTEx or CMC before those cohorts are described as independent evaluation datasets. This asymmetric, evaluation-cohort-dependent selection can inflate the reported coverage of trans models.
↳ Methods, “Cross-dataset training of INGENE and MODULE”; Results, “Evaluation of cis and trans models on a testing dataset”
The abstract associates improved imputation with 18,744 genes, whereas Figure 4a reports only 847–1,362 genes per region for which the cis–trans model was best. Uncorrected per-gene selection at α=0.05 further complicates interpretation of the pooled count.
↳ Abstract; Methods, “ANOVA analysis in GTEx”; Results, Figure 4a
The paper acknowledges that imputed-expression associations do not establish causality and that bulk tissue obscures cell types, yet concludes that network interactions contribute substantially to schizophrenia risk. The 641 newly reported associations also lack PGC-independent or functional validation.
↳ Discussion, limitations paragraph and final paragraph; Results, Figure 6
The paper makes a meaningful methodological contribution by adding two coexpression-informed trans predictors and applying the integrated framework at PGC3 scale. Its strongest execution features are the use of multiple brain cohorts, relevant cis and cis–trans comparators, and transparent reporting that cis models have higher external per-gene accuracy. The central uncertainty is created by cross-dataset filtering that uses GTEx- and CMC-trained trans models before those cohorts are presented as independent tests. Headline coverage, novelty, and mechanistic claims therefore require narrower wording and genuinely independent validation.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.0
Confidence mediumThe study adds two coexpression-informed trans predictors, compares them with established cis and cis–trans methods, and applies the integrated framework to a large schizophrenia dataset. The advance is meaningful, although it primarily expands predictive coverage rather than improving per-gene trans-model accuracy.
“both trans models substantially increased the number of predictable genes relative to cis-only approaches”
The analysis includes multiple cohorts, cross-validation, several baselines, and CMC evaluation of the combined model. However, trans-gene retention uses models trained in GTEx or CMC before those cohorts are described as independent tests, and no network-specific ablation isolates the contribution of coexpression structure.
“Genes were retained if at least one cross-dataset correlation was positive”
The manuscript follows a coherent progression from model construction through validation and coTWAS application, and it reports the lower external per-gene accuracy of trans predictors. The abstract’s claim of improved imputation for 18,744 genes is broader than Figure 4a’s much smaller per-region counts for which cis–trans was the best model.
“18,744 genes were significantly predicted by the integration of all models”
The paper engages with cis-TWAS and prior cis–trans frameworks and identifies limitations involving sample size, bulk tissue, ancestry, sex, and noncausal interpretation. The final mechanistic language and cell-type discussion do not fully preserve those limits.
“associations reflect genetically predicted expression and do not establish causality”
Lower confidence on Methodological Rigour — domain match limited.
Concerns4 of 4 checks
The purportedly independent GTEx and CMC evaluations follow a cross-dataset filtering step informed by models trained in those cohorts, making independence of the central coverage estimates materially uncertain. Several numerical and methodological statements also conflict within the supplied text.
The study declares informed consent and relevant institutional approvals for the source cohorts. Data, code, competing-interest, and related statements are deferred to online content not supplied here and are therefore assessability gaps, not conduct concerns.
Flags: 1 declared / 5 total
122 references in manuscript 114 of 118 checkable references found in an index 4 are books, websites or datasets — counted, but not index-checkable
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The framework has external computational evaluation and a large PGC3 application, but its outputs remain candidate associations for functional follow-up. No experimental validation, clinical implementation, or deployment-ready tool is demonstrated.
“associations reflect genetically predicted expression and do not establish causality”
AI-generated, human-governed. Something look off? Contact us to request a review.