Psychiatric genome-wide association study analyses implicate neuronal, immune and histone pathways

May combine two claims.

Synapse-related as well as newly implicated histone methylation and immune and neuronal signaling pathways are statistically significantly associated within and across schizophrenia, bipolar disorder and major depressive disorder.

“We show that synapse-related as well as newly implicated histone methylation and immune and neuronal signaling pathways are statistically significantly associated within and across SCZ, BIP and MDD.”

May combine two claims.

The pathways separate in multidimensional space and reveal two distinct branches for neuronal synapse- and histone methylation-related gene sets, with a third branch containing pathways with genes sharing immune or neurotrophic functional annotation.

“The pathways separate in multidimensional space and reveal two distinct branches for neuronal synapse– and histone methylation– related gene sets, with a third branch containing pathways with genes that share membership in pathways with immune or neurotrophic functional annotation (Fig. 3).”

May combine two claims.

The yellow module is similarly expressed across regions and contains half, 19 of 37, of the coexpressed histone methylation genes.

“Of note, the yellow module is similarly expressed across regions and contains half (19/37) of the coexpressed histone methylation genes.”

May combine two claims.

Two modules containing immune-neuronal signaling and synapse genes that plateau in expression in maturing brain exhibited cell type enrichment, with the brown module enriched for striatal neurons, particularly Drd1+ medium spiny neurons, and the turquoise module enriched for Cnp+ myelinating oligodendrocytes.

“Two of the modules containing immune-neuronal signaling and synapse genes that plateau in expression in maturing brain also exhibited cell type enrichment, with the brown module enriched for striatal neurons, particularly Drd1+ medium spiny neurons, and the turquoise module enriched for Cnp+ myelinating oligodendrocytes,”

• AssumptionRests on an assumption: Mouse cell type gene lists can be used to infer human cell type–specific patterns.

May combine two claims.

The null and HIV plots showed no enrichment of P values, indicating lack of shared biology and no systematic inflation of significance.

“The null and HIV plots shows no enrichment of P values, indicating both that there is a lack of shared biology (as expected) and that our analysis does not cause a systematic inflation of significance.”

• AssumptionThe authors state this as an assumption: HIV does not share significant disorder etiology with psychiatric disorders.

May combine two claims.

Of the 49 pathways, 16 are significant at q < 0.05, with the top pathway GO:51568 histone H3-K4 methylation having q = 0.0003.

“Of these, 16 are significant at q < 0.05, with the top pathway (GO:51568: histone H3-K4 methylation) having a q = 0.0003 (Table 2).”

May combine two claims.

There is statistically significant evidence for associations with multiple immune and neuronal signaling pathways and with the postsynaptic density.

“we also found statistically significant evidence for associations with multiple immune and neuronal signaling pathways and with the postsynaptic density”

May combine two claims.

The degree of rank correlation between pathways across pairs of disorders was significant, with 49 pathways with combined q-value <0.1 spanning the three adult disorders.

“The degree of rank correlation between pathways across pairs of disorders was significant (Supplementary Tables 4 and 6), with 49 pathways with combined q-value <0.1 spanning the three adult disorders (Table 2, Supplementary Table 7).”

May combine two claims.

We identified 10, 1 and 4 pathways suggestively enriched at FDR q-value <0.1 with bipolar disorder, schizophrenia and major depressive disorder susceptibility alleles, respectively.

“We identified 10, 1 and 4 pathway(s) that are suggestively enriched (at an FDR q-value <0.1) with BIP, SCZ and MDD susceptibility alleles, respectively (Table 1).”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The most significant pathway from the previous PGC Cross Disorder SNP-based meta-analysis, GO:0005262 calcium channel activity, showed nominally significant association in our pathway-level meta-analysis across all five disorders (P = 3.13 × 10−3) and also in schizophrenia, major depressive disorder and bipolar disorder (P = 8.07 × 10−3).

“the most significant pathway (GO:0005262, calcium channel activity) from the previous PGC Cross Disorder group SNP-based meta-analysis GWAS3 across the five disorders was found to show nominally significant association in our pathway level meta-analysis across all five disorders (P = 3.13 × 10−3) and also in SCZ, MDD and BIP (P = 8.07 × 10−3).”

• cites [3] Cross-Disorder Group of the Psychiatric Genomics Consortium et al. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379 (2013). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

No pathways showed significant enrichment in the null data set or HIV data set, and the schizophrenia, HIV and null data sets combined analysis gave no significant pathways after multiple-testing correction, with minimum q-value = 0.454.

“As expected, no pathways showed significant enrichment in the null data set or HIV data set, and the SCZ, HIV and null data sets combined analysis (Supplementary Table 8) gave no significant pathways after multiple-testing correction (minimum q-value = 0.454), in contrast to the results for SCZ, MDD and BIP shown in Table 2.”

• AssumptionThe authors state this as an assumption: HIV does not share significant disorder etiology with psychiatric disorders.

May combine two claims.

The strongest correlations were observed between the most similar methods: FORGE and SETSCREEN, and ALIGATOR and INRICH.

“As expected, the strongest correlations were observed between the most similar methods: FORGE and SETSCREEN, and ALIGATOR and INRICH.”

May combine two claims.

Expression patterns of coexpressed gene modules are predominantly driven by temporal regulation, with four modules showing greater than twofold temporal change and only one module showing a twofold change between regions.

“These expression patterns are predominantly driven by temporal regulation (four modules have greater than twofold temporal change, while only one module, green, shows a twofold change between regions, Supplementary Table 9)”

• AssumptionRests on an assumption: Gene expression data spanning brain regions and developmental time points can be used to place pathways in a specific neurobiological context.

May combine two claims.

Among individual disorders, bipolar disorder, schizophrenia, major depressive disorder and ADHD gave significant results at FDR q < 10%, but autism-spectrum disorder and HIV did not.

“Among individual disorders, BIP, SCZ, MDD and ADHD gave significant results at FDR q < 10%, but ASD and HIV did not (Table 1 and Supplementary Table 4).”

May combine two claims.

In the all-five-disease combination, GO:51568 stayed significant at q = 0.045, but its significance dropped because ADHD and ASD showed no enrichment.

“When all five diseases were combined, GO:51568 was still significant (q = 0.045), although its significance was reduced due to lack of enrichment in ADHD or ASD.”

May combine two claims.

Combining the three adult diseases yielded 15 pathways significant at q < 0.05, and GO:51568 had a q-value of 0.0005.

“When the three adult diseases were combined, 15 pathways were significant at q < 0.05, with the top pathway (GO:51568: histone H3-K4 methylation) having a q-value of 0.0005.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Kruskal-Wallis and Mann-Whitney U-tests show that many smaller spatial and temporal differences are statistically significant.

“though statistical analysis of spatial and temporal patterns with Kruskal-Wallis tests and pairwise Mann-Whitney U-tests demonstrates many smaller differences are significant (due to the large sample size).”

May combine two claims.

All pairwise correlations among the five diseases (ADHD, ASD, BIP, MDD, SCZ) were nominally significant, whereas correlations between these diseases and HIV or the null data set were not, with the exception of HIV-MDD, which was nominally significant but probably a multiple-testing artifact.

“Correlation coefficients in the actual data were nominally significant for all pairs of the five diseases of interest (ADHD, ASD, BIP, MDD, SCZ), but not between these diseases and HIV or the null data set (with the exception of the HIV-MDD correlation, which was nominally significant but is likely to be an artifact of multiple testing).”

Probably not a substantial claim: may not pass S3 (load-bearing).

Other approaches to assessing regional and temporal differences produced patterns similar to those in Figure 4b.

“In general, alternative methods for evaluating regional and temporal differences (for example, correlations with regional indicator variables, ANOVA with regions as factors) yielded similar patterns as those seen in Figure 4b.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Varying the WGCNA parameters still yielded similar coexpression modules.

“Similar modules were found with variations on these parameters.”

May combine two claims.

No pathway reached q < 0.05 in the remaining diseases or in the null data set.

“No pathway achieved q < 0.05 in the other diseases (or the null data set).”

Probably not a substantial claim: may not pass S3 (load-bearing).

Using the average of the best four ranks in a secondary analysis barely changed the findings.

“However, this had little effect on the results (see Supplementary Analysis and Supplementary Tables 12 and 13).”

May combine two claims.

Strong independent support of our findings comes from the observation that analyses of rare variants have also implicated postsynaptic pathways in schizophrenia etiology, suggesting that both rare and common variants are relevant.

“We note that strong, independent support of our findings comes from the observation that analyses of rare variants have also implicated postsynaptic pathways in SCZ etiology23,24, suggesting that both rare and common variants are relevant.”

• cites [23] Kirov, G. et al. De novo CNV analysis implicates specific abnormalities of postsynaptic signalling complexes in the pathogenesis of schizophrenia. Mol. Psychiatry 17, 142–153 (2012). sourceNot checked yet

  Not checked in the free run.

• cites [24] Purcell, S.M. et al. A polygenic burden of rare disruptive mutations in schizophrenia. Nature 506, 185–190 (2014). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

The most significant pathway from the PGC cross-disorder analysis, GO:0005262 calcium channel activity, also showed nominally significant enrichment in our analysis, confirming a role for calcium channel activity in these disorders.

“We note that the most significant pathway identified from the PGC cross-disorder analysis3 (GO:0005262, calcium channel activity) also showed nominally significant enrichment in our analysis, confirming a role for calcium channel activity in these disorders.”

• cites [3] Cross-Disorder Group of the Psychiatric Genomics Consortium et al. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379 (2013). sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Activity of voltage-gated calcium ions was implicated in a pathway analysis of a bipolar disorder GWAS data set.

“Activity of voltage-gated calcium ions was also implicated in a pathway analysis of a bipolar disorder GWAS data set11”

• cites [11] Psychiatric GWAS Consortium Bipolar Disorder Working Group. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat. Genet. 43, 977–983 (2011). sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Genome-wide association studies of psychiatric disorders have found multiple genetic associations with those disorders.

“Genome-wide association studies (GWAS) of psychiatric disorders have identified multiple genetic associations with such disorders”

May combine two claims.

Histone methylation appears to play a more prominent role in bipolar disorder, and synapse- and postsynapse-related processes are more strongly implicated in the etiology of schizophrenia.

“Our results suggest that histone methylation appears to play a more prominent role in bipolar disorder and that synapse- and postsynapse-related processes are more strongly implicated in the etiology of schizophrenia (Table 1).”

Classified as probably not a substantial claim.

Genetic pathway analysis methods for GWAS data aim to identify which biological pathways show an excess of etiological association.

“Genetic pathway analysis methods for GWAS data have been developed5,6 and aim to identify which biological pathways show an excess of etiological association”

• cites [5] Nam, D. & Kim, S.Y. Gene-set approach for expression pattern analysis. Brief. Bioinform. 9, 189–197 (2008). sourceNot checked yet

  Not checked in the free run.

• cites [6] Ackermann, M. & Strimmer, K. A general modular framework for gene set enrichment analysis. BMC Bioinformatics 10, 47 (2009). sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The lack of strong cell type specificity in other modules suggests that either they affect multiple cell lineages or the matching cell type profile has yet to be defined.

“suggesting that either they affect multiple cell lineages (for example, the yellow module) or the matching cell type profile has yet to be defined”

May combine two claims.

The pathway analyses described here do not require the same SNP or gene within a pathway to be implicated across disorders, so they are more robust to allelic heterogeneity within genes and within pathways across disorders.

“Importantly, the pathway analyses described here do not require the same SNP (or, indeed, gene) within a pathway to be implicated across disorders, so they will be more robust to allelic heterogeneity within genes and within pathways across disorders,”

May combine two claims.

Pathway analysis can translate GWAS signals into biochemical and/or system-wide understanding and can provide successful replication in the presence of allelic and locus heterogeneity.

“pathway analysis can translate GWAS signals into a level of understanding that is biochemical and/or system-wide and can provide successful replication in the presence of allelic and locus heterogeneity7,8”

• cites [7] Baranzini, S.E. et al. Pathway and network-based analysis of genome-wide association studies in multiple sclerosis. Hum. Mol. Genet. 18, 2078–2090 (2009). sourceNot checked yet

  Not checked in the free run.

• cites [8] Wang, K. et al. Diverse genome-wide association studies associate the IL12/IL23 pathway with Crohn Disease. Am. J. Hum. Genet. 84, 399–405 (2009). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

These pathways are core molecular processes, disruption of which may increase risk for multiple psychiatric disorders.

“These pathways are core molecular processes, disruption of which may increase risk for multiple psychiatric disorders.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The cell-type enrichment analysis applied multiple-comparison correction across seven modules and 35 cell-type gene lists.

“We assessed each module for enrichment of lists for 35 broad and specific cell type gene sets (CSEA specificity threshold set to 0.05) across multiple brain regions in mouse, and corrected for multiple comparisons across 7 modules and the 35 cell type lists.”

Classified as probably not a substantial claim.

Every P value underwent genomic-control correction.

“Empirical results (P values for individual SNPs) from PGC GWAS were gathered for each disorder, and all P values were GC-corrected.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The combined P-value method ignores dependence among pathways, so it cannot assess whether significant pathways are over-represented, though it remains valid for testing each pathway individually.

“Note that this procedure does not allow for dependence between pathways, so it cannot be used to test whether there is an excess of pathways with average rank achieving a given level of significance. However, it does give a valid test of significance for each pathway separately.”

Classified as probably not a substantial claim.

Raw genotype and phenotype data for each study were uploaded to a central server and processed through the same quality control, imputation and analysis pipeline to ensure comparability across samples.

“To ensure comparability across samples, raw genotype and phenotype data for each study were uploaded to a central server and processed through the same quality control, imputation and analysis pipeline.”

May combine two claims.

Given the role of histone methylation in establishing active promoters, it appears likely that dysregulation of histone methylation may have downstream effects with the potential to disrupt neurodevelopment and coordinated gene expression, as animal studies have demonstrated.

“Given the role of this process in establishing active promoters38, it appears likely that the dysregulation of histone methylation may have downstream effects with the potential to disrupt neurodevelopment and coordinated gene expression, as animal studies have demonstrated39.”

• cites [38] Brykczynska, U. et al. Repressive and active histone methylation mark distinct promoters in human and mouse spermatozoa. Nat. Struct. Mol. Biol. 17, 679–687 (2010). sourceNot checked yet

  Not checked in the free run.

• cites [39] Huang, H.S. et al. Prefrontal dysfunction in schizophrenia involves mixed-lineage leukemia 1-regulated histone methylation at GABAergic gene promoters. J. Neurosci. 27, 11254–62 (2007). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

A pathway with a lower average rank shows more consistent enrichment across methods and is more likely to contribute to disease susceptibility.

“The lower the average rank of a pathway, the more consistent its evidence for enrichment of association signal across the methods, and thus the greater the likelihood of involvement in disease susceptibility.”

• AssumptionRests on an assumption: A pathway that truly contributes to disease susceptibility should show consistent enrichment of association signal across multiple analysis methods.

Classified as probably not a substantial claim.

Method-overlap comparisons used only independent pathways, obtained by Jaccard-distance pruning.

“To robustly test the significance of overlap in enriched pathways between methods, it is necessary to restrict the analysis to a set of independent pathways (by gene membership). This was done by pruning by Jaccard distance (see below).”

Classified as probably not a substantial claim.

Further analyses were restricted to 4,949 gene sets of at most 200 genes and at least 10 genes to ensure specificity of gene set association findings.

“To ensure the specificity of gene set association findings, further analyses were restricted to 4,949 gene sets of at most 200 genes and at least 10 genes”

Probably not a substantial claim: may not pass S3 (load-bearing).

The MHC region on chromosome 6, 25-35 Mbp, was excluded from further analyses to prevent potential impact of extensive linkage disequilibrium in the region.

“The MHC region on chromosome 6 (25–35 Mbp) was excluded from further analyses to prevent potential impact of extensive linkage disequilibrium (LD) in the region.”

Classified as probably not a substantial claim.

Transcriptional regulatory elements are likely located within 35 kb upstream and 10 kb downstream of genes.

“since transcriptional regulatory elements are likely to be contained within these intervals and that there is thus merit in capturing the variation within these regions42.”

• cites [42] Maston, G.A., Evans, S.K. & Green, M.R. Transcriptional regulatory elements in the human genome. Annu. Rev. Genomics Hum. Genet. 7, 29–59 (2006). sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

A null GWAS sample was analyzed to assess the degree of dependence between methods and to ensure that the analysis did not generate an excessive amount of false positives.

“We also analyzed a null GWAS sample, to assess the degree of dependence between methods, as well as to ensure that our analysis did not generate an excesive amount of false positives.”

Probably not a substantial claim: may not pass S3 (load-bearing).

SNPs with imputation quality INFO ≤ 0.8 were excluded.

“SNPs were also filtered by imputation quality (INFO > 0.8), which resulted in 477,792–543,578 SNPs being assigned to 16,334–17,352 genes”

Classified as probably not a substantial claim.

Stringent and standardized quality control procedures were applied as previously described.

“Stringent and standardized quality control procedures were applied as previously described3.”

• cites [3] Cross-Disorder Group of the Psychiatric Genomics Consortium et al. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379 (2013). sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

Reported enrichments use a Benjamini-Hochberg corrected significance threshold of P < 0.05.

“We report enrichments at Benjamini-Hochberg corrected P < 0.05 in Figure 4c.”

May combine two claims.

Our autism-spectrum disorder data set did not have sufficient power and was excluded from the primary pathway analysis.

“However, our ASD data set, based on GWAS, did not have sufficient power (owing to both the smaller sample size available and the combination of case-control and trio data sets) and was excluded from the primary pathway analysis.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

INRICH, ALIGATOR, and MAGENTA are best-SNP-per-gene methods that count pathway genes with independent SNPs above a significance threshold, correcting for LD and genomic structure via Monte-Carlo simulation.

“INRICH45, ALIGATOR9 and MAGENTA46 are “best SNP per gene” methods that count the number of genes in a pathway where a number of independent SNPs exceed a predefined significance, and adjust for LD and genomic structure with corrected statistics derived by Monte-Carlo simulation.”

• cites [45] Lee, P.H., O'Dushlaine, C., Thomas, B. & Purcell, S.M. INRICH: interval-based enrichment analysis for genome-wide association studies. Bioinformatics 28, 1797–1799 (2012). sourceNot checked yet

  Not checked in the free run.

• cites [9] Holmans, P. et al. Gene ontology analysis of GWA study data sets provides insights into the biology of bipolar disorder. Am. J. Hum. Genet. 85, 13–24 (2009). sourceNot checked yet

  Not checked in the free run.

• cites [46] Segrè, A.V., Groop, L., Mootha, V.K., Daly, M.J. & Altshuler, D. Common inherited variation in mitochondrial genes is not enriched for associations with type 2 diabetes or related glycemic traits. PLoS Genet. 6, e1001058 (2010). sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

These results should be considered preliminary, as they are a first-pass analysis using the results of a novel approach.

“We emphasize that these results should be considered preliminary, as they are a first-pass analysis using the results of a novel approach.”

Classified as probably not a substantial claim.

The analysis selected 1,918 pathways with pairwise Jaccard similarity not exceeding 0.2.

“a subset of 1,918 pathways was selected such that no two pathways had a Jaccard similarity measure >0.2.”

Probably not a substantial claim: may not pass S3 (load-bearing).

SNPs with imputation quality scores less than 0.8 were filtered out.

“SNPs with imputation quality scores less than 0.8 were filtered out.”

Classified as probably not a substantial claim. May combine two claims.

FORGE and SET SCREEN TEST are meta-analytic approaches that pool SNP P values within genes or pathways while correcting for linkage disequilibrium.

“FORGE43 and SET SCREEN TEST44 are meta-analysis methods that combine P values across all the SNPs in genes or pathways while adjusting for the confound of LD.”

• cites [43] Pedroso, I. et al. Common genetic variants and gene-expression changes associated with bipolar disorder are over-represented in brain signaling pathway genes. Biol. Psychiatry 72, 311–317 (2012). sourceNot checked yet

  Not checked in the free run.

• cites [44] Moskvina, V. et al. Evaluation of an approximation method for assessment of overall significance of multiple-dependent tests in a genomewide association study. Genet. Epidemiol. 35, 861–866 (2011). sourceNot checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

The analysis was repeated with and without the MHC region included.

“Analysis was run both with and without the MHC region (chromosome 6, 25–35 Mb).”

May combine two claims.

The degree of correlation between pathways across disorders is higher than expected by chance, indicating that polygenic overlap is nonrandom at a molecular or pathway level.

“Our observation that the degree of correlation between pathways across disorders is higher than expected by chance builds on the observation of shared genetics between these disorders4 and, importantly, indicates that polygenic overlap is nonrandom at a molecular or pathway level.”

• cites [4] Mirnics, K., Middleton, F.A., Marquez, A., Lewis, D.A. & Levitt, P. Molecular characterization of schizophrenia viewed by microarray analysis of gene expression in prefrontal cortex. Neuron 28, 53–67 (2000). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

Combining the three adult diseases produced stronger significance than any single-disease analysis, demonstrating the power gained by pooling pathway analyses across diseases that share biology.

“These results are more significant than those observed in any of the diseases analyzed separately, and illustrate the power to be gained by combining pathway analyses across diseases with shared biology.”

Classified as probably not a substantial claim. May combine two claims.

Only one of each pair of related or duplicate individuals was retained, and in only one disorder case or control set, to ensure independence of individual disorder analyses.

“to ensure independence of individual disorder analyses, only one of pair of related or duplicate individuals were retained, and in only one disorder case or control set, resulting in 61,220 cases and controls in total.”

Classified as probably not a substantial claim.

Pathway multiple-testing correction was performed by calculating q-values.

“To correct for multiple testing of pathways, q-values were calculated48.”

• cites [48] Storey, J.D. & Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. USA 100, 9440–9445 (2003). sourceNot checked yet

  Not checked in the free run.

May combine two claims.

The blue module shows an increase during late prenatal and early postnatal development, peaking before age 6, suggesting the genes in this module may be related to postnatal synaptic pruning, another potential target for developing interventions.

“It shows an increase during late prenatal and early postnatal development, peaking before age 6, suggesting the genes in this module may be related to postnatal synaptic pruning, another potential target for developing interventions.”

May combine two claims.

FORGE and SET-SCREEN may produce extremely small enrichment P values for pathways with strongly associated SNPs, unlike ALIGATOR, INRICH, and MAGENTA, which rely on simulation.

“Specifically, FORGE and SET-SCREEN combine the P values of the SNPs, and it is therefore possible for them to achieve very small enrichment P values if the pathway contains strongly associated SNPs. This is not the case for ALIGATOR, INRICH and MAGENTA, which use simulation to give enrichment P values.”