CoChat
Sign Up
Cited by 965Genomics and Chromatin DynamicsGenetic Associations and EpidemiologySingle-cell and spatial transcriptomics

Comprehensive functional genomic resource and integrative model for the human brain

Open access
Wang, Daifeng, Liu, Shuang, Warrell, Jonathan, Won, Hyejung, Shi, Xu, Navarro, Fabio C. P., Clarke, Declan, Gu, Mengting +186 more·2018·Science
DOI 10.1126/science.aat8464PMC PMC6413328
Open this paper in CoChat — free account

Abstract

INTRODUCTION: Strong genetic associations have been found for a number of psychiatric disorders. However, understanding the underlying molecular mechanisms remains challenging. RATIONALE: To address this challenge, the PsychENCODE Consortium has developed a comprehensive online resource and integrative models for the functional genomics of the human brain. RESULTS: The base of the pyramidal resource is the datasets generated by PsychENCODE, including bulk transcriptome, chromatin, genotype, and Hi-C datasets and single-cell transcriptomic data from ~32,000 cells for major brain regions. We have merged these with data from Genotype-Tissue Expression (GTEx), ENCODE, Roadmap Epigenomics, and single-cell analyses. Via uniform processing, we created a harmonized resource, allowing us to survey functional genomics data on the brain over a sample size of 1866 individuals. From this uniformly processed dataset, we created derived data products. These include lists of brain-expressed genes, coexpression modules, and single-cell expression profiles for many brain cell types; ~79,000 brain-active enhancers with associated Hi-C loops and topologically associating domains; and ~2.5 million expression quantitative-trait loci (QTLs) comprising ~238,000 linkage-disequilibrium–independent single-nucleotide polymorphisms and of other types of QTLs associated with splice isoforms, cell fractions, and chromatin activity. By using these, we found that >88% of the cross-population variation in brain gene expression can be accounted for by cell fraction changes. Furthermore, a number of disorders and aging are associated with changes in cell-type proportions. The derived data also enable comparison between the brain and other tissues. In particular, by using spectral analyses, we found that the brain has distinct expression and epigenetic patterns, including a greater extent of noncoding transcription than other tissues. The top level of the resource consists of integrative networks for regulation and machine-learning models for disease prediction. The networks include a full gene regulatory network (GRN) for the brain, linking transcription factors, enhancers, and target genes from merging of the QTLs, generalized element-activity correlations, and Hi-C data. By using this network, we link disease genes to genome-wide association study (GWAS) variants for psychiatric disorders. For schizophrenia, we linked 321 genes to the 142 reported GWAS loci. We then embedded the regulatory network into a deep-learning model to predict psychiatric phenotypes from genotype and expression. Our model gives a ~6-fold improvement in prediction over additive polygenic risk scores. Moreover, it achieves a ~3-fold improvement over additive models, even when the gene expression data are imputed, highlighting the value of having just a small amount of transcriptome data for disease prediction. Lastly, it highlights key genes and pathways associated with disorder prediction, including immunological, synaptic, and metabolic pathways, recapitulating de novo results from more targeted analyses. CONCLUSION: Our resource and integrative analyses have uncovered genomic elements and networks in the brain, which in turn have provided insight into the molecular mechanisms underlying psychiatric disorders. Our deep-learning model improves disease risk prediction over traditional approaches and can be extended with additional data types (e.g., microRNA and neuroimaging). A comprehensive functional genomic resource for the adult human brain.: The resource forms a three-layer pyramid. The bottom layer includes sequencing datasets for traits, such as schizophrenia. The middle layer represents derived datasets, including functional genomic elements and QTLs. The top layer contains integrated models, which link genotypes to phenotypes. DSPN, Deep Structured Phenotype Network; PC1 and PC2, principal components 1 and 2; ref, reference; alt, alternate; H3K27ac, histone H3 acetylation at lysine 27.

AI snapshot

Written from the paper’s full text.

TL;DR

The PsychENCODE Consortium assembled an adult brain functional genomics resource spanning 1,866 prefrontal cortex individuals, integrating transcriptomic, epigenomic, single-cell, Hi-C, and QTL data. Embedding the resulting regulatory network in an interpretable deep-learning model improved psychiatric trait prediction by about sixfold over polygenic risk scores and linked schizophrenia GWAS loci to candidate genes.

Key findings

  • The resource contains 3,810 PsychENCODE datasets plus 1,662 external bulk datasets, with prefrontal cortex sampling from 1,866 individuals; single-cell RNA-seq covers 18,025 PsychENCODE cells and 14,012 external cells. It annotates 79,056 prefrontal cortex enhancers, 2,735 topologically associating domains, and about 90,000 enhancer-promoter interactions.
  • Cell-type proportion variation explains more than 88% of cross-population bulk expression variation; estimated NEU+/− fractions matched reference samples with a median difference of 0.04. Cell fraction shifts were associated with sex, autism spectrum disorder (higher Ex5, lower oligodendrocytes), and age (increases in Ex3 and Ex4).
  • QTL analyses identified about 2.5 million cis-eQTLs involving roughly 33,000 eGenes, including 1,341,182 eQTL SNPs from about 5.3 million tested and 238,194 independent SNPs after LD pruning. Additional findings include about 2,000 cQTLs plus 6,200 from CMC, 1,672 SNPs forming 4,199 fQTLs, 202,940 SNPs in residual eQTLs, and 2,477 multi-QTLs.
  • The regulatory network linked 142 schizophrenia GWAS loci to 1,111 putative schizophrenia genes across 119 loci, with 321 high-confidence genes. The deep-learning model improved prediction by about sixfold (+22.9% versus +3.8% for genome alone) and achieved about a 3.1-fold improvement over logistic prediction for genotype-only inference; liability-scale variance explained was 32.8% for schizophrenia and 37.4% for bipolar disorder, but 11.3% for autism spectrum disorder.

Methods

  • The study built a hierarchical resource by merging PsychENCODE adult brain data with ENCODE, CommonMind Consortium, GTEx, Roadmap, and single-cell datasets, processing RNA-seq, ChIP-seq, DNase-seq, ATAC-seq, and Hi-C through standardized pipelines.
  • Analyses included bulk and single-cell transcriptome profiling, nonnegative matrix factorization, deconvolution of bulk expression into cell-type signatures and fractions, enhancer annotation using H3K27ac and H3K4me3, QTL mapping for expression, isoforms, chromatin, cell fractions, and residual expression, Hi-C-based regulatory linkage, transcription-factor target regression, GWAS variant-to-gene mapping, and a Deep Structured Phenotype Network.
  • Key sample sizes include 1,866 prefrontal cortex individuals, 50 individuals for H3K27ac across prefrontal cortex, temporal cortex, and cerebellum, 292 individuals for chromatin QTLs, and 18,025 plus 14,012 single cells from PsychENCODE and external sources.

Limitations

  • The paper states that single-cell methods have limited sensitivity and missing values, making low-abundance transcripts and fine subcellular structures hard to quantify; it also assumes cell-type signature expression is relatively constant across individuals.
  • The paper notes that cQTLs rely on one-dimensional chromatin signal and may miss three-dimensional architecture, that larger cohorts and additional chromatin assays are needed, and that the deep-learning model may help mainly polygenic traits rather than Mendelian or population-structure-correlated traits; liability estimates may be confounded by environmental variation.
  • Evident from the text: raw genotype and sequencing data are restricted for privacy and require approval; the H3K27ac cohort of 50 individuals was primarily of European descent; enhancer variability across individuals and tissues is high, with only about 70% ± 15% of reference enhancers active per individual and saturation estimated after about 30 individuals.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1126/science.aat8464

Cites

109 of 109 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 74 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The PFC datasets involve sampling from 1866 individuals.

    “Overall, the datasets from the prefrontal cortex (PFC) involve sampling from 1866 individuals.”
    Quote in the paperNot a substantial claim
  • Findingnone given

    All data were placed into a central publicly available resource that also integrates reprocessed data from ENCODE, CMC, GTEx, and Roadmap.

    “All data have been placed into a central, publicly available resource that also integrates relevant reprocessed data from related projects, including ENCODE, the CommonMind Consortium (CMC), GTEx, and Roadmap.”
    Quote in the paperSubstantial claim
  • Findingcitation

    PsychENCODE generated large-scale data including genotyping, bulk and single-cell RNA-seq, ChIP-seq, ATAC-seq, and Hi-C.

    “the PsychENCODE Consortium has generated large-scale data to provide insight into the brain and psychiatric disorders, including data derived through genotyping, bulk and single-cell RNA sequencing (RNA-seq), chromatin immunoprecipitation with sequencing (ChIP-seq), assay for transposase-accessible chromatin using sequencing (ATAC-seq), and Hi-C (17).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [17] Akbarian S, The PsychENCODE project. Nat. Neurosci 18, 1707– 1712 (2015). doi: 10.1038/nn.4156; pmid: 26605881 26605881 PMC4675669 source

      Not checked in the free run.

  • Findingown data

    The resource data represent a range of psychiatric disorders including SCZ, BPD, and ASD.

    “These data represent a range of psychiatric disorders, including SCZ, bipolar disorder (BPD), and autism spectrum disorder (ASD).”
    Quote in the paperNot a substantial claim
  • Findingown data

    The PsychENCODE Consortium produced a comprehensive online resource for adult brain covering 1866 individuals.

    “the PsychENCODE Consortium has generated a comprehensive online resource for the adult brain across 1866 individuals.”
    Quote in the paperSubstantial claim
  • May combine two claims. 44
  • Classified as probably not a substantial claim. 10
  • The quote only partly states this claim. 1

  • May combine two claims.

    Findingcitation

    The resource has single-cell RNA-seq data for 18,025 cells from PsychENCODE and 14,012 cells from outside sources.

    “The resource also has single-cell RNA-seq data for 18,025 cells from PsychENCODE and 14,012 cells from outside sources (20).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [20] Lake BB, Integrative single-cell analysis of transcriptional and epigenetic states in the human adult brain. Nat. Biotechnol 36, 70– 80 (2018). doi: 10.1038/nbt.4038; pmid: 29227469 29227469 PMC5951394 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    The resource enables construction of a gene regulatory network linking GWAS variants to genes, including 321 genes for schizophrenia.

    “It also allows building of a gene regulatory network, linking genome-wide association study variants to genes (e.g., 321 for schizophrenia).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Many GWAS variants associated with neuropsychiatric disease lie in noncoding regions.

  • May combine two claims.

    Findingown data

    Integration of the resource shows that variation in cell-type proportions largely explains cross-population expression variation, with over 88% reconstruction accuracy.

    “Integration shows that varying cell-type proportions largely account for the cross-population variation in expression (with >88% reconstruction accuracy).”
    • Also stated in: § Transcriptome analysis: Bulk and single cell · hedged
      “weighted combinations of single-cell signatures could account for most of the population-level expression variation, with an accuracy of >88% (Fig. 2D)”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Brain tissue is composed of a variety of basic cell types.

  • May combine two claims.

    Findingown data

    The authors identified noncoding regions of transcription and sets of differentially expressed and coexpressed genes.

    “Using these data, we identified noncoding regions of transcription and sets of differentially expressed and coexpressed genes (21, 22).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

    • cites [22] Gandal MJ, Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362, eaat8127 (2018). 30545856 PMC6443102 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Embedding the network into an interpretable deep-learning model improves disease prediction by about sixfold compared with polygenic risk scores and identifies key genes and pathways in psychiatric disorders.

    “We embed this network into an interpretable deep-learning model, which improves disease prediction by ~6-fold versus polygenic risk scores and identifies key genes and pathways in psychiatric disorders.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Many GWAS variants associated with neuropsychiatric disease lie in noncoding regions.

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

    Findingown data

    The cell-type profiles were in two formats, TPM and UMI counts.

    “Depending on the underlying sequencing and quantification, our profiles were of two fundamentally different formats, transcripts per kilobase million (TPM) and unique molecular identifier (UMI) counts.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    On average, only about 70% ± 15% (about 54,000) of reference enhancers were active in an individual in the cohort, and 68% were active in more than half the cohort.

    “On average, only ~70% ± 15% (~54,000) of the enhancers in the reference brain were active in an individual in the cohort, and a similar fraction of the reference enhancers was active in more than half the cohort (68%) (Fig. 3B).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    FZD9 expression levels were associated with a neighboring noncoding SNP via an eQTL, and the same SNP was associated with the proportion of Ex3 cells via an fQTL.

    “We found that the expression levels of this gene were associated with a neighboring noncoding SNP via an eQTL, and this same SNP was associated with the proportion of Ex3 cells via an fQTL.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    eQTLs tended to be enriched at promoters, cQTLs at enhancers and TF-binding sites, and fQTLs were spread over many different elements.

    “As expected, eQTLs tended to be enriched at promoters, and cQTLs, at enhancers and transcription factor (TF)–binding sites; fQTLs were spread over many different elements.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Aggregating data across the cohort resulted in 37,761 H3K27ac peaks in PFC, 42,683 in TC, and 26,631 in CB, each present in more than half of individuals.

    “Aggregating data across the cohort resulted in a total of 37,761 H3K27ac “peaks” (enriched regions) in the PFC, 42,683 in the TC, and 26,631 in the CB—where each peak is present in more than half of the individuals surveyed.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The authors found 1,341,182 eQTL SNPs from about 5.3 million total SNPs tested in 1-Mb windows, constituting 238,194 independent SNPs after LD pruning.

    “We found 1,341,182 eQTL single-nucleotide polymorphisms (SNPs) from ~5.3 million total SNPs tested in 1-Mb windows around genes, constituting 238,194 independent SNPs after linkage-disequilibrium (LD) pruning.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    In individuals with ASD, the fraction of Ex5 was higher and that of oligodendrocytes lower, with some increase for microglia and astrocytes.

    “in individuals with ASD, the fraction of Ex5 was higher and that of oligodendrocytes, lower, with some commensurate increase for microglia and astrocytes (Fig. 2E and fig. S18) (24, 29).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [24] Gandal MJ, Shared molecular neuropathology across major psychiatric disorders parallels polygenic overlap. Science 359, 693– 697 (2018). doi: 10.1126/science.aad6469; pmid: 29439242 29439242 PMC5898828 source

      Not checked in the free run.

    • cites [29] Rubenstein JL, Merzenich MM, Model of autism: Increased ratio of excitation/inhibition in key neural systems. Genes Brain Behav. 2, 255– 267 (2003). doi: 10.1034/j.1601-183X.2003.00037.x; pmid: 14606691 14606691 PMC6748642 source

      Not checked in the free run.

    • Rests on an assumption: Brain tissue is composed of a variety of basic cell types.

  • May combine two claims.

    Findingown data

    Weighted combinations of single-cell signatures account for most population-level expression variation, with accuracy over 88%.

    “weighted combinations of single-cell signatures could account for most of the population-level expression variation, with an accuracy of >88% (Fig. 2D) (1 − ||B − CW||2/||B||2 > 88%)”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Brain tissue is composed of a variety of basic cell types.

  • The quote only partly states this claim.

    Findingown data

    The top NMF components of bulk expression variation were consistently associated with single-cell signatures.

    “we decomposed the bulk gene expression matrix by using nonnegative matrix factorization (NMF) (B ≈ VH, where B, V, and H represent matrices) and determined whether the top components (NMF-TCs), capturing the majority of covariance (columns of V) (Fig. 2B), were consistently associated with the single-cell signatures (Fig. 2C) (21).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingown data

    In the PFC, the authors identified about 2.5 million cis-eQTLs involving about 33,000 eGenes, with about 17,000 noncoding and about 16,000 coding, at FDR < 0.05.

    “In the PFC, we identified ~2.5 million cis-eQTLs involving ~33,000 eGenes (expressed genes) [~17,000 noncoding and ~16,000 coding, with a false discovery rate of <0.05] (Fig. 4A).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The authors identified 2735 TADs and about 90,000 enhancer-promoter interactions.

    “In total, we identified 2735 topologically associating domains (TADs) and ~90,000 enhancer-promoter interactions (fig. S36).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    With increasing age, fractions Ex3 and Ex4 significantly increased and fractions of some nonneuronal types decreased.

    “with increasing age, the fractions Ex3 and Ex4 significantly increased and the fractions of some nonneuronal types decreased (Fig. 2F and fig. S19).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Brain tissue is composed of a variety of basic cell types.

  • May combine two claims.

    Findingcitation

    Fetus-linked genes were highly expressed in adult neurons, whereas adult-linked ones were preferentially expressed in glia.

    “They were also highly expressed in adult neurons, whereas the adult-linked ones were preferentially expressed in glia, reflecting known cell-type composition (Fig. 5, D and F) (42).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42] de la Torre-Ubieta L, Won H, Stein JL, Geschwind DH, Advancing the understanding of autism disease mechanisms through genetics. Nat. Med 22, 345– 361 (2016). doi: 10.1038/nm.4071; pmid: 27050589 27050589 PMC5072455 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Adult brain Hi-C datasets clustered together, while the fetal brain Hi-C dataset was distinct.

    “whereas Hi-C datasets for the adult brain clustered together, the Hi-C dataset for the fetal brain was distinct (Fig. 5D and fig. S39).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Fetus-linked genes were more highly expressed prenatally and adult-linked ones postnatally.

    “we found fetus-linked genes to be more highly expressed prenatally and adult-linked ones postnatally (Fig. 5E).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    About 75% of enhancer-promoter interactions occurred within the same TAD, and genes with more enhancers tended to have higher expression.

    “As expected, ~75% of enhancer-promoter interactions occurred within the same TAD, and genes with more enhancers tended to have higher expression (Fig. 5B and fig. S36).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The cumulative number of active regions increased for the first 20 individuals sampled but saturated at the 30th.

    “This increased for the first 20 individuals sampled but saturated at the 30th.”
    Quote in the paperSubstantial claim

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

    Findingown data

    The authors provide an expanded regulatory network with slightly different parameterization.

    “Lastly, for broad utility on the resource website, we also provide an expanded regulatory network with slightly different parameterization (fig. S42).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The authors developed a comprehensive deep-learning model, the DSPN, and used it to illustrate how functional genomics data could improve the link between genotype and phenotype.

    “Third, we developed a comprehensive deep-learning model, the DSPN, and used it to illustrate how functional genomics data could improve the link between genotype and phenotype.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The full interaction between genotype and phenotype involves many levels beyond those encapsulated by the regulatory network.

  • May combine two claims.

    Findingcitation

    The authors take an unbiased view of all 5024 modules and higher-order groupings and prioritize a subset of ~180 modules and groupings for each psychiatric disorder, showing these to be enriched in specific functional categories and to intersect substantially with modules from more disease-focused analyses.

    “Broadly, we take an unbiased view of all 5024 modules and higher-order groupings constructed from these and then prioritize a subset of ~180 modules and groupings for each psychiatric disorder, showing these to be enriched in specific functional categories and to intersect substantially with the modules from more disease-focused analyses (Fig. 8, B and C, and fig. S49) (22).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [22] Gandal MJ, Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362, eaat8127 (2018). 30545856 PMC6443102 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    The 1111 schizophrenia genes (and the 321 high-confidence subset) resembled known schizophrenia-associated genes in several ways, showing enrichment for translational regulators, cholinergic receptors, calcium channels, synaptic genes, schizophrenia differentially expressed genes, and loss-of-function-intolerant genes.

    “As expected, they shared many characteristics with known SCZ-associated genes, being enriched in translational regulators, cholinergic receptors, calcium channels, synaptic genes, SCZ differentially expressed genes, and loss-of-function-intolerant genes (fig. S45) (47).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [47] Pardiñas AF, Common schizophrenia alleles are enriched in mutation-intolerant genes and in regions under strong background selection. Nat. Genet 50, 381– 389 (2018). doi: 10.1038/s41588-018-0059-2; pmid: 29483656 29483656 PMC5918692 source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    The schizophrenia genes represent an increase from the 22 genes reported in an earlier QTL study and a larger number than can be linked by genomic proximity (176).

    “Overall, the SCZ genes represent an increase from the 22 genes reported in an earlier QTL study and a larger number than can be linked simply by genomic proximity (176) (Fig. 6A) (11, 47).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [11] Fromer M, Gene expression elucidates functional impact of polygenic risk for schizophrenia. Nat. Neurosci 19, 1442– 1453 (2016). doi: 10.1038/nn.4399; pmid: 27668389 27668389 PMC5083142 source

      Not checked in the free run.

    • cites [47] Pardiñas AF, Common schizophrenia alleles are enriched in mutation-intolerant genes and in regions under strong background selection. Nat. Genet 50, 381– 389 (2018). doi: 10.1038/s41588-018-0059-2; pmid: 29483656 29483656 PMC5918692 source

      Not checked in the free run.

  • May combine two claims.

    Findingreasoning

    By using a generative architecture, the DSPN can impute intermediate phenotypes and provide forward predictions from genotypes to traits.

    “By using a generative architecture, we ensure that the model is able to impute intermediate phenotypes, as well as provide forward predictions from genotypes to traits.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    The authors identified a set of eQTLs several times as large as those in previous studies, targeting a saturating proportion of protein-coding genes.

    “First, in terms of QTLs, we identified a set of eQTLs several times as large as those in previous studies, targeting a saturating proportion of protein-coding genes.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The variance explained by the full DSPN model exceeds that explained by common SNPs in schizophrenia and bipolar disorder (32.8 and 37.4% respectively; autism spectrum disorder 11.3% slightly lower).

    “By contrast, the variance explained by the full DSPN model exceeds that explained by common SNPs in SCZ and BPD, possibly reflecting the influence of rare variants and epistatic interactions (32.8 and 37.4% respectively—the variance of 11.3% for ASD is slightly lower).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    Other prioritized modules include well-characterized genes such as MIAT, RBFOX1, and ANK2 (schizophrenia); RELA, NFkB2, and NIPBL (autism); and HOMER1 (bipolar disorder).

    “Other prioritized modules include well-characterized genes such as MIAT, RBFOX1, and ANK2 (SCZ); RELA, NFkB2, and NIPBL (ASD); and HOMER1 (BPD), consistent with the results of (22).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] Gandal MJ, Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362, eaat8127 (2018). 30545856 PMC6443102 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Modules associated with aging are enriched in Ex4 neuronal cell-type genes, synaptic and longevity functions, and the gene NRGN.

    “Finally, we identify modules associated with aging, finding that they are enriched in Ex4 neuronal cell-type genes, synaptic and longevity functions, and the gene NRGN—all consistent with differential expression analysis (Fig. 8D and fig. S20).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The DSPN includes layers for intermediate molecular phenotypes, predefined gene groupings, higher layers for inferred groupings, and a top layer for observed traits.

    “We included layers for intermediate molecular phenotypes associated with specific genes (i.e., their gene expression and chromatin state) and predefined gene groupings (cell-type marker genes and coexpression modules), multiple higher layers for inferred groupings (hidden nodes), and a top layer for observed traits (psychiatric disorders and other brain phenotypes).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The authors determined which modules were prioritized and the sets of genes associated with latent nodes at each hidden layer.

    “Using it, we determined which modules were prioritized, as well as the sets of genes associated with latent nodes that were found at each hidden layer (Fig. 8A and table S10) (15, 21).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [15] Saliba AE, Westermann AJ, Gorski SA, Vogel J, Single-cell RNA-seq: Advances and future challenges. Nucleic Acids Res. 42, 8845– 8860 (2014). doi: 10.1093/nar/gku555; pmid: 25053837 25053837 PMC4132710 source

      Not checked in the free run.

    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Findingown data

    The DSPN allows joint inference and imputation of intermediate phenotypes and observed traits from genotype alone, achieving a ~3.1-fold improvement over a logistic predictor.

    “Moreover, the DSPN allows us to perform joint inference and imputation of intermediate phenotypes (i.e., transcriptome and epigenome) and observed traits from just the genotype alone, achieving a ~3.1-fold improvement over a logistic predictor in this context (Fig. 7C and fig. S46).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    For schizophrenia, the authors find enrichments for pathways and genes known to be associated with the disease, including glutamatergic-synapse pathway genes such as GRIN1; calcium-signaling pathway and astrocyte-marker genes; and complement cascade pathway genes such as C4A, C4B, and CLU.

    “For SCZ, which is the best characterized of the three disorders, we find enrichments for pathways and genes known to be associated with the disease, including glutamatergic-synapse pathway genes, such as GRIN1; calcium-signaling pathway and astrocyte-marker genes; and complement cascade pathway genes such as C4A, C4B, and CLU (Fig. 8D) (22).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [22] Gandal MJ, Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362, eaat8127 (2018). 30545856 PMC6443102 source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductcitation

    PsychENCODE developed a reference brain project on the PFC using matched assays on the same brain tissues, which was used to develop an anchoring annotation.

    “PsychENCODE has developed a reference brain project on the PFC by using matched assays on the same set of brain tissues, which we used to develop an anchoring annotation (21).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Classified as probably not a substantial claim.

    Premisecitation

    ENCODE and Roadmap Epigenomics identified noncoding regions acting as enhancers and promoters.

    “the ENCODE and Roadmap Epigenomics (Roadmap) projects have identified noncoding regions acting as enhancers and promoters (8–10).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [8] Battle A, Genetic effects on gene expression across human tissues. Nature 550, 204– 213 (2017). doi: 10.1038/nature24277; pmid: 29022597 29022597 PMC5776756 source

      Not checked in the free run.

    • cites [9] Kundaje A, Integrative analysis of 111 reference human epigenomes. Nature 518, 317– 330 (2015). doi: 10.1038/nature14248; pmid: 25693563 25693563 PMC4530010 source

      Not checked in the free run.

    • cites [10] ENCODE Project Consortium, An integrated encyclopedia of DNA elements in the human genome. Nature 489, 57– 74 (2012). doi: 10.1038/nature11247; pmid: 22955616 22955616 PMC3439153 source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductreasoning

    RNA-seq data from PsychENCODE, GTEx, and Roadmap were uniformly processed with the standardized ENCODE pipeline, making expression data and subsequent results comparable with previous work.

    “we took a conservative approach and used the standardized and established ENCODE pipeline to uniformly process RNA-seq data from PsychENCODE, GTEx, and Roadmap (figs. S2 and S3).”
    • Also stated in: § Transcriptome analysis: Bulk and single cell · plain
      “This consistency makes our expression data and subsequent results (including eQTLs and single-cell analyses) comparable with previous work.”
    Quote in the paperSubstantial claim

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

    Premisecitation

    GTEx has associated many noncoding variants with expression quantitative-trait loci.

    “the Genotype-Tissue Expression (GTEx) project has associated many noncoding variants with expression quantitative-trait loci (eQTLs)”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [8] Battle A, Genetic effects on gene expression across human tissues. Nature 550, 204– 213 (2017). doi: 10.1038/nature24277; pmid: 29022597 29022597 PMC5776756 source

      Not checked in the free run.

    • cites [9] Kundaje A, Integrative analysis of 111 reference human epigenomes. Nature 518, 317– 330 (2015). doi: 10.1038/nature14248; pmid: 25693563 25693563 PMC4530010 source

      Not checked in the free run.

    • cites [10] ENCODE Project Consortium, An integrated encyclopedia of DNA elements in the human genome. Nature 489, 57– 74 (2012). doi: 10.1038/nature11247; pmid: 22955616 22955616 PMC3439153 source

      Not checked in the free run.

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

    Premisecitation

    The Psychiatric Genomics Consortium has identified 142 GWAS loci associated with schizophrenia.

    “the Psychiatric Genomics Consortium (PGC) has identified 142 GWAS loci associated with schizophrenia (SCZ) (6)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [6] Geschwind DH, Flint J, Genetics and genomics of psychiatric disease. Science 349, 1489– 1494 (2015). doi: 10.1126/science.aaa8954; pmid: 26404826 26404826 PMC4694563 source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    Both TPM and UMI profiles share common neuronal and major nonneuronal cell types and are used interchangeably in analyses.

    “Both formats share common neuronal and major nonneuronal cell types and are used interchangeably in various analyses in this study (fig. S5 and tables S3 and S4).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Initial brain-specific functional genomics work provided greater insight but could be enhanced with larger sample sizes.

    “Initial work focusing on brain-specific functional genomics has provided greater insight but could be enhanced with larger sample sizes (11, 12).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [11] Fromer M, Gene expression elucidates functional impact of polygenic risk for schizophrenia. Nat. Neurosci 19, 1442– 1453 (2016). doi: 10.1038/nn.4399; pmid: 27668389 27668389 PMC5083142 source

      Not checked in the free run.

    • cites [12] Colantuoni C, Temporal dynamics and genetic control of transcription in the human prefrontal cortex. Nature 478, 519– 523 (2011). doi: 10.1038/nature10524; pmid: 22031444 22031444 PMC3510670 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    RCA captures local structure while maintaining meaningful distances globally.

    “RCA is a compromise (21): It captures local structure while maintaining meaningful distances globally.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Study design and conductcitation

    H3K27ac data from PFC, TC, and CB were uniformly processed on a cohort of 50 individuals, primarily of European descent and sequenced to similar depths.

    “we uniformly processed the H3K27ac data from the PFC, temporal cortex (TC), and cerebellum (CB) on a cohort of 50 individuals, primarily of European descent and sequenced to similar depths (21) (fig. S24).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • Probably not a substantial claim: may not pass S1 (asserted).

    Interpretationown data

    Pooling PFC enhancers from about 30 individuals is sufficient to cover nearly all possible PFC enhancer regions, estimated at about 120,000.

    “Thus, we hypothesize that pooling PFC enhancers from ~30 individuals is sufficient to cover nearly all possible PFC enhancer regions, estimated at ~120,000.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    The difference in similarity between PFC/TC and CB is consistent with previous reports and suggests potentially different cell-type composition in CB and cortex.

    “This difference is consistent with previous reports and suggests potentially different cell-type composition in the CB and the cortex (33, 34).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [33] Sun W, Histone acetylome-wide association study of autism spectrum disorder. Cell 167, 1385– 1397.e11 (2016). doi: 10.1016/j.cell.2016.10.031; pmid: 27863250 27863250 source

      Not checked in the free run.

    • cites [34] Purves D, Neuroscience (Oxford Univ., ed. 6, 2018).

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationown data

    This estimate identified substantially more eQTLs and associated eGenes than previous studies, reflecting the large sample size.

    “This estimate identified substantially more eQTLs and associated eGenes than previous studies, reflecting our large sample size (8, 11, 21).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [8] Battle A, Genetic effects on gene expression across human tissues. Nature 550, 204– 213 (2017). doi: 10.1038/nature24277; pmid: 29022597 29022597 PMC5776756 source

      Not checked in the free run.

    • cites [11] Fromer M, Gene expression elucidates functional impact of polygenic risk for schizophrenia. Nat. Neurosci 19, 1442– 1453 (2016). doi: 10.1038/nn.4399; pmid: 27668389 27668389 PMC5083142 source

      Not checked in the free run.

    • cites [21] Materials and methods are available as supplementary materials.

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

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

    Premisecitation

    Deletion variants upstream of FZD9 had previously been associated with cell fraction changes related to Williams syndrome.

    “deletion variants upstream of FZD9 had previously been associated with cell fraction changes related to Williams syndrome (40).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [40] Chailangkarn T, A human neurodevelopmental model for Williams syndrome. Nature 536, 338– 343 (2016). doi: 10.1038/nature19067; pmid: 27509850 27509850 PMC4995142 source

      Not checked in the free run.

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

    Premisecitation

    LHX9 is critical for early cortical specification and SOX7 for neuronal apoptosis.

    “These include LHX9 and SOX7, TFs critical for early cortical specification and neuronal apoptosis, respectively (48, 49).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [48] Mangale VS, Lhx2 selector activity specifies cortical identity and suppresses hippocampal organizer fate. Science 319, 304– 309 (2008). doi: 10.1126/science.1151695; pmid: 18202285 18202285 PMC2494603 source

      Not checked in the free run.

    • cites [49] Wang C, SOX7 interferes with β-catenin activity to promote neuronal apoptosis. Eur. J. Neurosci 41, 1430– 1437 (2015). doi: 10.1111/ejn.12910; pmid: 25847511 25847511 source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The authors kept their eQTL methods very standard, closely following the GTEx paradigm.

    “In this study, we kept our eQTL methods very standard, closely following the GTEx paradigm.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationown data

    Additional data would be beneficial.

    “Our best liability scores (from just the genotype at QTL-associated variants) are substantially below these bounds, implying that additional data would be beneficial.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    The authors expect the DSPN will improve accuracy mainly for complex traits with a highly polygenic architecture, but not necessarily for traits strongly determined by few variants or closely correlated with population structure.

    “We expect that the DSPN will improve accuracy mainly for complex traits with a highly polygenic architecture, but not necessarily for traits that are strongly determined by only a few variants, such as Mendelian disorders, or are closely correlated with population structure, such as ethnicity.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    The transcriptome carries additional information, which the DSPN is able to extract.

    “This result clearly manifests that the transcriptome carries additional information, which the DSPN is able to extract.”
    Quote in the paperSubstantial claim
54 of 54

  • The paper relies on this without support.

    It is unclear how changes in cell proportions contribute to variation in tissue-level gene expression across a population.

    “However, it is unclear how these changes in cell proportions can contribute to the variation in tissue-level gene expression observed across a population of individuals.”

  • The paper relies on this without support.

    None of GTEx, ENCODE, and Roadmap focused on the human brain.

    “However, none of these projects have focused their efforts on the human brain.”

  • The paper relies on this without support.

    Many GWAS variants associated with neuropsychiatric disease lie in noncoding regions.

    “Many of these variants lie in noncoding regions”

  • The authors state this as an assumption.

    Elastic net regression assumes target gene expression is determined by a linear combination of expression levels of regulating TFs via regression coefficients.

    “Elastic net regression assumes that target gene expression is determined by a linear combination of the expression levels of its regulating TFs, via regression coefficients (using sparsified L1 and L2 regularization).”

  • The paper relies on this without support.

    Brain tissue is composed of a variety of basic cell types.

    “Brain tissue is composed of a variety of basic cell types.”

  • The paper relies on this without support.

    The full interaction between genotype and phenotype involves many levels beyond those encapsulated by the regulatory network.

    “The full interaction between genotype and phenotype involves many levels, beyond those encapsulated by the regulatory network.”

  • The authors state this as an assumption.

    The expression levels characterizing a signature are fairly constant over a population of cells of a given cell type.

    “However, this assumes that the expression levels characterizing a signature are fairly constant over a population of cells of a given cell type.”

  • The paper relies on this without support.

    Traditional classification methods such as logistic regression predict phenotype directly from genotype, without using intermediates such as the transcriptome.

    “Traditional classification methods such as logistic regression predict phenotype directly from genotype, without using intermediates such as the transcriptome (Fig. 7A).”
8 of 8

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Cell type annotations were supported by expression patterns of curated marker genes.

    “Annotations were supported by expression patterns of curated marker genes16,17 (Extended Data Fig. 1e; Supplementary Table S2).”

    From APOE4 impairs myelination via cholesterol dysregulation in oligodendrocytes (2022) · cites this paper as [16]

    In “Transcriptome analysis: Bulk and single cell” (page not provided), the paper describes brain cell-type profiles and expression signatures comprising marker genes, and reports that biomarker-gene expression for the same cell type correlated across two data formats. Our copy does not say the marker genes were curated or that their expression patterns were used to support the cell-type annotations. (LLM, conf=0.78)

    Experimental — accuracy not yet measured

  • miR-QTLs were enriched in PsychEncode bulk prefrontal cortex enhancers (Mantel-Haenszel common odds ratio = 1.32, 95% CI = [1.20,1.45], p = 1.3 × 10−8).

    “miR-QTLs were enriched in this set of enhancers (Mantel-Haenszel common odds ratio = 1.32, 95% CI = [1.20,1.45], p = 1.3 × 10−8)”

    From Mapping the microRNA landscape in the older adult brain and its genetic contribution to neuropsychiatric conditions (2024) · cites this paper as [32]

    The Enhancers section (page unavailable) identifies a reference set of 79,056 prefrontal cortex enhancers, supporting the identification of the enhancer set. The QTL analysis section discusses enrichment of other QTL types at genomic elements, but our copy does not report miR-QTL enrichment or the stated odds ratio, confidence interval, and p-value. (LLM, conf=0.88)

    Experimental — accuracy not yet measured

  • Heritability for all psychiatric disorders and behavioral traits was strongly enriched in neuronal enhancers and promoters, and much less so in PsychENCODE bulk brain enhancers.

    “We found a strong enrichment of heritability for variants within neuronal enhancers and promoters for all psychiatric disorders and behavioral traits (Fig. 1D), which was substantially lower in PsychENCODE bulk brain enhancers (Fig. 1D) (13).”

    From Brain cell type–specific enhancer–promoter interactome maps and disease <b>-</b> risk association (2019) · cites this paper as [13]

  • Promoters of genes that mark a given cell type tended to show the matching H3K27ac profile.

    “Promoters associated with cell type signature genes preferentially exhibited corresponding H3K27ac profiles”

    From Brain cell type–specific enhancer–promoter interactome maps and disease <b>-</b> risk association (2019) · cites this paper as [13]

    The Transcriptome analysis: Bulk and single cell section describes cell-type expression signatures and marker genes, while the Enhancers section describes H3K27ac profiles. Neither passage reports whether promoters of those marker genes preferentially exhibit H3K27ac profiles corresponding to the same cell type. I searched for the stated promoter–signature-gene relationship and cell-type-specific promoter acetylation; page numbers were not available in our copy. (LLM, conf=0.73)

    Experimental — accuracy not yet measured

  • Ninety-four percent of bulk brain enhancers from PsychENCODE overlapped with enhancers found in the sorted nuclei.

    “Most bulk brain enhancer regions identified by PsychENCODE overlapped with the nuclei cell type enhancers (94%) (13).”

    From Brain cell type–specific enhancer–promoter interactome maps and disease <b>-</b> risk association (2019) · cites this paper as [13]

    The Enhancers section describes a reference set of 79,056 brain enhancers, but it does not report their overlap with enhancers from sorted nuclei. Searches for “94%,” “sorted nuclei enhancers,” and “cell type enhancers overlap bulk” found no passage establishing the claimed comparison. (LLM, conf=0.72)

    Experimental — accuracy not yet measured

Cites

Flagged first, then by citations

Research works printed but not linked: 4

  1. D. Purves Neuroscience (Oxford Univ. ed. 6 2018).
  2. R. Salakhutdinov G. Hinton “Deep Boltzmann machines ” Proc. 12th Int. Conf. Artif. Intell. Stat. 5 448–455 (2009).
  3. K. Simonyan Ph.D. dissertation University of Oxford (2013).
  4. J. Grove S. Ripke T. Damm Als M. Mattheisen R. Walters H. Won J. Pallesen E. Agerbo O. A. Andreassen R. Anney R. Belliveau F. Bettella J. D. Buxbaum J. Bybjerg-Grauholm M. Bækved-Hansen F. Cerrato K. Chambert J. H. Christensen C. Churchhouse K. Dellenvall D. Demontis S. De Rubeis B. Devlin S. Djurovic A. Dumont J. Goldstein C. S. Hansen M. E. Hauberg M. V. Hollegaard S. Hope D. P. Howrigan H. Huang C. Hultman L. Klei J. Maller J. Martin A. R. Martin J. Moran M. Nyegaard T. Nærland D. S. Palmer A. Palotie C. B. Pedersen M. G. Pedersen T. Poterba J. B. Poulsen B. St. Pourcain P. Qvist K. Rehnström A. Reichenberg J. Reichert E. Robinson K. Roeder P. Roussos E. Saemundsen S. Sandin F. K. Satterstrom G. D. Smith H. Stefansson K. Stefansson S. Steinberg C. Stevens P. F. Sullivan P. Turley G. B. Walters X. Xu ASD Working Group of the PGC BUPGEN Major Depressive Disorder Working Group of the PGC 23andMe Research Team D. Geschwind M. Nordentoft D. M. Hougaard T. Werge O. Mors P. B. Mortensen B. M. Neale M. J. Daly A. D. Børglum Common risk variants identified in autism spectrum disorder. bioRxiv 224774 [Preprint]. 27 November 2017. .10.1101/224774 · DOI · Paper page

Cited by

Sorted by citations

Related

Authors
Wang, D., Liu, S., Warrell, J., Won, H., Shi, X., Navarro, F. C. P., Clarke, D., Gu, M., Emani, P., Yang, Y. T., Xu, M., Gandal, M. J., Lou, S., Zhang, J., Park, J. J., Yan, C., Rhie, S. K., Manakongtreecheep, K., Zhou, H., Nathan, A., Peters, M., Mattei, E., Fitzgerald, D., Brunetti, T., Moore, J., Jiang, Y., Girdhar, K., Hoffman, G. E., Kalayci, S., Gümüş, Z. H., Crawford, G. E., PsychENCODE Consortium, Roussos, P., Akbarian, S., Jaffe, A. E., White, K. P., Weng, Z., Sestan, N., Geschwind, D. H., Knowles, J. A., Gerstein, M. B., Ashley-Koch, A. E., Crawford, G. E., Garrett, M. E., Song, L., Safi, A., Johnson, G. D., Wray, G. A., Reddy, T. E., Goes, F. S., Zandi, P., Bryois, J., Jaffe, A. E., Price, A. J., Ivanov, N. A., Collado-Torres, L., Hyde, T. M., Burke, E. E., Kleiman, J. E., Tao, R., Shin, J. H., Akbarian, S., Girdhar, K., Jiang, Y., Kundakovic, M., Brown, L., Kassim, B. S., Park, R. B., Wiseman, J. R., Zharovsky, E., Jacobov, R., Devillers, O., Flatow, E., Hoffman, G. E., Lipska, B. K., Lewis, D. A., Haroutunian, V., Hahn, C.-G., Charney, A. W., Dracheva, S., Kozlenkov, A., Belmont, J., DelValle, D., Francoeur, N., Hadjimichael, E., Pinto, D., van Bakel, H., Roussos, P., Fullard, J. F., Bendl, J., Hauberg, M. E., Mangravite, L. M., Peters, M. A., Chae, Y., Peng, J., Niu, M., Wang, X., Webster, M. J., Beach, T. G., Chen, C., Jiang, Y., Dai, R., Shieh, A. W., Liu, C., Grennan, K. S., Xia, Y., Vadukapuram, R., Wang, Y., Fitzgerald, D., Cheng, L., Brown, M., Brown, M., Brunetti, T., Goodman, T., Alsayed, M., Gandal, M. J., Geschwind, D. H., Won, H., Polioudakis, D., Wamsley, B., Yin, J., Hadzic, T., De La Torre Ubieta, L., Swarup, V., Sanders, S. J., State, M. W., Werling, D. M., An, J.-Y., Sheppard, B., Willsey, A. J., White, K. P., Ray, M., Giase, G., Kefi, A., Mattei, E., Purcaro, M., Weng, Z., Moore, J., Pratt, H., Huey, J., Borrman, T., Sullivan, P. F., Giusti-Rodriguez, P., Kim, Y., Sullivan, P., Szatkiewicz, J., Rhie, S. K., Armoskus, C., Camarena, A., Farnham, P. J., Spitsyna, V. N., Witt, H., Schreiner, S., Evgrafov, O. V., Knowles, J. A., Gerstein, M., Liu, S., Wang, D., Navarro, F. C. P., Warrell, J., Clarke, D., Emani, P. S., Gu, M., Shi, X., Xu, M., Yang, Y. T., Kitchen, R. R., Gürsoy, G., Zhang, J., Carlyle, B. C., Nairn, A. C., Li, M., Pochareddy, S., Sestan, N., Skarica, M., Li, Z., Sousa, A. M. M., Santpere, G., Choi, J., Zhu, Y., Gao, T., Miller, D. J., Cherskov, A., Yang, M., Amiri, A., Coppola, G., Mariani, J., Scuderi, S., Szekely, A., Vaccarino, F. M., Wu, F., Weissman, S., Roychowdhury, T., & Abyzov, A.
Affiliations
  • Program in Computational Biology and Bioinformatics, Yale University, New Haven, CT 06520, USA.
  • Yale University (US)
  • Sage Bionetworks, Seattle, WA 98109, USA.
  • Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA.
  • Duke University (US)
  • Sage Bionetworks (US)
  • Icahn School of Medicine at Mount Sinai (US)
  • University of Chicago (US)
  • University of California, Los Angeles (US)
ORCID
Venue
Science
Publisher
American Association for the Advancement of Science (AAAS)
Published
2018-12-14
Volume
362
Issue
6420
ISSN
  • 0036-8075 (ISSN-L)
  • 1095-9203
DOI
10.1126/science.aat8464
PMC
PMC6413328
OpenAlex
W2904763455
Semantic Scholar
4e256382e2064492e75427c83911feac91758671
Cited by
965
References
109
Citations by year
2026: 73 · 2025: 117 · 2024: 132 · 2023: 140 · 2022: 138 · 2021: 141 · 2020: 131 · 2019: 86 · 2018: 5 · 2017: 1
FWCI
34.67
Type
journal article
Language
en
Open access
Locations
  • Science (publishedVersion)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (acceptedVersion)
  • PubMed Central (acceptedVersion)
  • eScholarship@UMassChan (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
Funders
National Institute of Mental Health (R01 MH094714, R01 MH110927, R01 MH105472)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Chromatin Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
Keywords
PsychENCODE, mental illness, gene regulatory networks, expression quantitative trait loci, topologically associating domains, chromatin looping, brain cell types, schizophrenia, polygenic risk prediction, brain aging, enhancers, single-cell transcriptomics, chromatin accessibility, disease prediction model, brain-expressed genes, gene co-expression modules
MeSH terms
Deep Learning, Mental Disorders, Brain, Enhancer Elements, Genetic, Gene Expression Regulation, Humans, Quantitative Trait Loci, Epigenesis, Genetic, Gene Regulatory Networks, Genome-Wide Association Study, Epigenomics, Single-Cell Analysis, Transcriptome, Datasets as Topic

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-05.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-05.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review