CoChat
Sign Up
Cited by 290Genetic Associations and EpidemiologyGenetic Mapping and Diversity in Plants and AnimalsBioinformatics and Genomic Networks

Genomic atlas of the proteome from brain, CSF and plasma prioritizes proteins implicated in neurological disorders

Open access
Yang, Chengran, Farias, Fabiana H. G., Ibanez, Laura, Suhy, Adam, Sadler, Brooke, Fernandez, Maria Victoria, Wang, Fengxian, Bradley, Joseph L. +14 more·2021·Nature Neuroscience
DOI 10.1038/s41593-021-00886-6PMC PMC8521603

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Understanding the tissue-specific genetic controls of protein levels is essential to uncover mechanisms of post-transcriptional gene regulation. In this study, we generated a genomic atlas of protein levels in three tissues relevant to neurological disorders (brain, cerebrospinal fluid and plasma) by profiling thousands of proteins from participants with and without Alzheimer’s disease. We identified 274, 127 and 32 protein quantitative trait loci (pQTLs) for cerebrospinal fluid, plasma and brain, respectively. cis-pQTLs were more likely to be tissue shared, but trans-pQTLs tended to be tissue specific. Between 48.0% and 76.6% of pQTLs did not co-localize with expression, splicing, DNA methylation or histone acetylation QTLs. Using Mendelian randomization, we nominated proteins implicated in neurological diseases, including Alzheimer’s disease, Parkinson’s disease and stroke. This first multi-tissue study will be instrumental to map signals from genome-wide association studies onto functional genes, to discover pathways and to identify drug targets for neurological diseases. Yang et al. generated a genomic atlas of protein levels in brain, cerebrospinal fluid and plasma and used human genetics approaches to identify proteins implicated in neurological diseases as well as druggable targets.

AI snapshot

Written from the paper’s full text.

TL;DR

This study generated a multi-tissue protein quantitative trait locus (pQTL) atlas from brain, CSF, and plasma in participants with and without Alzheimer disease (AD). It identified tissue-shared and tissue-specific pQTLs, showed many pQTLs do not colocalize with other molecular QTLs, and used Mendelian randomization (MR) to prioritize proteins and drug targets for neurological disorders.

Key findings

  • The study identified 274 independent pQTLs for 184 CSF proteins, 127 for 100 plasma proteins, and 32 for 27 brain proteins. Cis associations accounted for 82% of CSF, 76% of plasma, and 94% of brain pQTLs; trans associations accounted for 18%, 24%, and 6%, respectively, and 91.5–98.9% of trans-pQTLs were inter-chromosomal.
  • Cis-pQTLs were more likely to be tissue-shared, whereas trans-pQTLs were more tissue-specific: 15–26% of cis-pQTLs were tissue-specific versus 77–91% of trans-pQTLs. CSF and plasma each shared more than 70% of brain pQTLs.
  • Between 48.0% (brain) and 76.6% (CSF) of pQTLs did not colocalize with expression, splicing, DNA-methylation, or histone-acetylation QTLs. Only 13.3–33.3% of pQTLs colocalized with GTEx cortex eQTLs and 14.2–17.5% with GTEx whole-blood eQTLs at PP.H4 > 0.8.
  • MR nominated proteins implicated in neurological diseases, including three CSF, 13 plasma, and seven brain proteins for AD risk, and proteins such as plasma CD33 for AD, plasma IDUA for Parkinson’s disease, CSF Carbonic Anhydrase IV for ALS, and CSF/plasma E-Selectin for stroke. The study identified 25 drugs known to interact with MR-prioritized proteins.

Methods

  • Aptamer-based proteomics measured 1,305 proteins in CSF (n = 971), plasma (n = 636), and brain parietal lobe cortex (n = 458) samples from participants of European ancestry, including AD cases and cognitively normal individuals. After QC, analyses included 835 CSF samples/713 proteins, 529 plasma samples/931 proteins, and 380 brain samples/1,079 proteins.
  • Genome-wide association analyses tested 14.06 million imputed autosomal common variants (MAF ≥ 0.02) against protein levels, with cis defined as within 1 Mb of the gene start site and trans as outside a 2 Mb window. Significance thresholds were p < 5×10−8 for cis-pQTLs and study-wide thresholds for trans-pQTLs: 2.96×10−10 for CSF, 2.17×10−10 for plasma, and 6.67×10−10 for brain.
  • The study used conditional analyses, replication and meta-analyses with independent datasets, mashr for tissue specificity, colocalization with eQTL/sQTL/mQTL/hQTL data, MR with TwoSampleMR, and DrugBank for drug repositioning.

Limitations

  • The paper states that the aptamer-based platform may be confounded by recognition differences rather than true protein abundance changes, and that coding-variant enrichment could be biased by the platform. It also notes that non-coding pQTLs may remain undetected due to limited statistical power, that colocalization results cannot be extrapolated to all coding genes, and that brain pQTLs came from parietal lobe cortex while GTEx eQTLs came from whole cortex.
  • The paper states that MR cannot fully account for horizontal pleiotropy, although pleiotropic regions were removed, and that protein–neurological disease associations may not be disease-specific, as associations with asthma risk were also observed. It calls for additional studies in more tissues and with more proteins.
  • Evident from the text: sample sizes and case/control composition differ substantially across tissues, and the brain control-only analysis had a much smaller sample size (24 controls), which the authors note produced a relatively low correlation.

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

Integrity signals

Registry

Crossref · 10.1038/s41593-021-00886-6

Cites

93 of 93 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 24 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    91.5% to 98.9% of trans-pQTLs were inter-chromosomal.

    “We found that 91.5 to 98.9% of the trans-pQTLs were inter-chromosomal (Table S6).”
    Quote in the paperNot a substantial claim
  • Findingnone given

    The number of significant pQTLs was proportional to sample size rather than the number of proteins.

    “The number of significant pQTLs was proportional to the sample size, rather than the number of proteins.”
    Quote in the paperNot a substantial claim
  • Findingown data

    In CSF, 82% of the 274 significant associations were cis and 18% were trans.

    “Of the 274 significant associations in CSF, 82% were cis associations and 18% were trans associations.”
    Quote in the paperNot a substantial claim
  • Findingown data

    In the second round of conditional analyses, 23 CSF, 6 plasma, and 1 brain loci still showed independent and genome-wide significant associations.

    “In the second round, 23 CSF, 6 plasma, and 1 brain loci still showed independent and genome-wide significant associations (Fig. 2a).”
    Quote in the paperNot a substantial claim
  • Findingown data

    In brain, 94% of significant associations were cis and 6% were trans.

    “Lastly, in brain, 94% were cis associations and 6% were trans associations (Fig. 1b,c).”
    Quote in the paperNot a substantial claim
  • May combine two claims. 30
  • Classified as probably not a substantial claim. 24

  • May combine two claims.

    Findingown data

    The study found 1 protein with up to 5 independent signals, 10 proteins with 4 independent signals, and 30 proteins with 3 independent signals.

    “We found 1 protein with up to 5 independent signals, 10 proteins with 4 independent signals, and 30 proteins with 3 independent signals.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    96.8% of plasma pQTLs and 96.9% of brain pQTLs were replicated.

    “This is supported by the fact that we have been able to replicate 96.8% and 96.9% of plasma and brain pQTLs (see below).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    CSF meta-analysis including 596 common proteins identified 425 pQTLs for 310 proteins, compared to 250 pQTLs for 185 proteins in our CSF cohort alone.

    “Due to the increased sample-size, we identified 425 pQTLs for 310 proteins, compared to 250 pQTLs for 185 proteins identified in our CSF cohort alone.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    After conditioning on the main CSF ARTS1 signal, a genome-wide signal tagged by rs467735 remained and was associated with gene expression.

    “After conditioning on this signal, there was still a genome-wide signal tagged by rs467735 (conditional p-value=9.82×10−89) that was associated with gene expression (GTEx; multi-tissue p-value=0).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For brain, 5 (15.6%) signals replicated at genome-wide significance, 8 (25%) at nominal association, and 18 (56.3%) showed at least nominal association in brain or CSF; only 1 (3.1%) was not replicated.

    “We were able to replicate 5 (15.6%) signals at genome-wide significance, 8 (25%) signals at a nominal (p-value < 0.05) association, and 18 (56.3%) pQTLs that showed at least a nominal association in brain or CSF. Only 1 (3.1%) pQTL was not replicated.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    There were an additional 32 pleiotropic regions in plasma and 9 in brain.

    “We found an additional 32 pleiotropic regions in plasma and 9 in brain.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    In meta-analysis of the PPMI and Sasayama studies, 153 (55.8%) of our significant signals had genome-wide significance and 27 (9.9%) additional pQTLs showed at least nominal association in the same direction.

    “We found that 153 (55.8%) of our significant signals had genome-wide significance in our meta-analysis of the PPMI19 and Sasayama studies and 27 (9.9%) additional pQTLs showed at least a nominal association (p-value < 0.05) in the same direction.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    CD33 is a target of AVE9633, an anti-CD33 antibody used to treat acute myeloid leukemia.

    “Our analyses also indicated that CD33 is a target of AVE9633, an anti-CD33 antibody used to treat acute myeloid leukemia, and therefore this antibody could be used in the context of AD.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Mendelian randomization identified three proteins involved in Alzheimer disease risk in CSF, 13 in plasma, and 7 in brain.

    “After multiple testing correction, we identified three proteins involved in AD risk in CSF, 13 in plasma, and 7 in brain.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Only 13.3% to 33.3% of pQTLs had a posterior probability of hypothesis-4 (PP.H4)>0.8 for GTEx cortex eQTLs or 14.2% to 17.5% for GTEx whole blood eQTLs.

    “Colocalization analyses indicated that only 13.3 to 33.3% of pQTLs had a posterior probability of hypothesis-4 (PP.H4)>0.8 for the GTEx cortex eQTLs or 14.2 to 17.5% for GTEx whole blood eQTLs (Tables S18, S19).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    42.5% of protein-disease associations colocalized with GWAS loci, including plasma Siglec-3 (CD33) and Alzheimer disease risk.

    “We found that 42.5% of protein-disease associations colocalized with GWAS loci, including plasma Siglec-3 (also known as CD33) and AD risk (Tables S29, S30).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    A SNP at the IDUA locus (rs35220088) is associated with plasma IDUA protein level and Parkinson's disease risk (p=2.47×10−6 and 3.52×10−9, respectively).

    “We found a SNP at this locus that is associated with plasma IDUA protein level and PD risk (rs35220088; p= 2.47×10−6 and 3.52×10−9, respectively).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    83% to 90% of non-coding pQTLs and 10% to 17% of coding pQTLs were also seen in a mass spectrometry-based pQTL study.

    “However, we found 83 to 90% of our non-coding pQTL and 10 to 17% coding pQTL, respectively and depending on tissues, were also seen in a mass spectrometry-based pQTL study16 (Table S32).”
    Quote in the paperSubstantial claim

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

    Findingown data

    The average time difference between the two CSF LP dates was 6.14 years (SD 2.98).

    “The average time difference between the two LP dates was 6.14 years, and the standard deviation was 2.98 years.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The correlation between default model and controls-only model was lower in brain (0.602) than in CSF (0.999) and plasma (0.998).

    “Using model 3 for comparison, we observed a Pearson correlation coefficient of 0.999, 0.998, 0.602 (p-value = 0.002) for CSF, plasma, and brain, respectively.”
    Quote in the paperNot a substantial claim

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

    Findingown data

    The estimates of variance explained by the top variant were similar to using the top variant as the only predictor after accounting for covariates.

    “To account for variance explained by covariates, we also calculated the adjusted R-squared value using a nested model and the estimates were similar to using the top variant as the only predictor (Supplementary Fig. 4).”
    Quote in the paperNot a substantial claim

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

    Findingown data

    For CSF, the variance explained by conditionally independent variants had mean = 0.141, standard deviation = 0.144, and mode = 0.061.

    “For CSF, the mean = 0.141, standard deviation = 0.144, mode = 0.061; For plasma, the mean = 0.157, standard deviation = 0.125, mode = 0.188; For brain, the mean = 0.208, standard deviation = 0.151, mode = 0.092.”
    Quote in the paperNot a substantial claim

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

    Findingown data

    For plasma, the variance explained by conditionally independent variants had mean = 0.157, standard deviation = 0.125, and mode = 0.188.

    “For CSF, the mean = 0.141, standard deviation = 0.144, mode = 0.061; For plasma, the mean = 0.157, standard deviation = 0.125, mode = 0.188; For brain, the mean = 0.208, standard deviation = 0.151, mode = 0.092.”
    Quote in the paperNot a substantial claim

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

    Findingown data

    For brain, the variance explained by conditionally independent variants had mean = 0.208, standard deviation = 0.151, and mode = 0.092.

    “For CSF, the mean = 0.141, standard deviation = 0.144, mode = 0.061; For plasma, the mean = 0.157, standard deviation = 0.125, mode = 0.188; For brain, the mean = 0.208, standard deviation = 0.151, mode = 0.092.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    E-Selectin is a known stroke biomarker.

    “E-Selectin is a known stroke biomarker24, and our MR plus colocalization analyses indicated that this protein is genetically associated with the disease (Tables S27, S30).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] Ibanez, L. et al. Overlap in the genetic architecture of stroke risk, early neurological changes, and cardiovascular risk factors. Stroke 50, 1339–1345 (2019). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    CSF Siglec-9, plasma sICAM-1, and brain PPAC have been replicated in other studies using a different proteomic approach, indicating these findings may not be driven by platform.

    “CSF Siglec-9, plasma sICAM-1, and brain PPAC have been replicated in other studies16,29,31,32 using a different proteomic approach, which indicates that these findings may not be driven by platform.”
    Quote in the paperNot a substantial claim0 of 4 sources checked
    • cites [16] Robins, C. et al. Genetic control of the human brain proteome. Preprint at bioRxiv https://doi.org/10.1101/816652 (2019). source

      Not checked in the free run.

    • cites [29] Hillary, R. F. et al. Genome and epigenome wide studies of neurological protein biomarkers in the Lothian Birth Cohort 1936. Nat. Commun. 10, 3160–3160 (2019). source

      Not checked in the free run.

    • cites [31] Yao, C. et al. Genome‐wide mapping of plasma protein QTLs identifies putatively causal genes and pathways for cardiovascular disease. Nat. Commun. 9, 3268 (2018). source

      Not checked in the free run.

    • cites [32] Paré, G. et al. Novel association of ABO histo-blood group antigen with soluble ICAM-1: results of a genome-wide association study of 6,578 women. PLoS Genet. 4, e1000118 (2008). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The analysis validated CD33 as a key protein to Alzheimer disease; it is also part of the TREM2-MS4A4A pathway.

    “In summary, this analysis validated CD33 as a key protein to AD; it is also part of the TREM2-MS4A4A pathway.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Acetazolamide is known to inhibit carbonic anhydrases.

    “Acetazolamide, used to treat glaucoma, epilepsy, and altitude sickness, is known to inhibit carbonic anhydrases39.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [39] Bethea, J. W. Clinical Anesthesia, 6th Edition. Anesthesiology 112, 767–768 (2010). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    CD33 likely affects microglia activity and innate immune response.

    “CD33 is a microglia-specific gene36 and likely affects microglia activity and innate immune response.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    The data suggest that around half of pQTLs do not colocalize with other molecular QTLs, but this data cannot be extrapolated to all coding genes.

    “Our data suggest that around half of pQTLs do not colocalize with other molecular QTLs, but this data cannot be extrapolated to all coding genes.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    This is the first study to address the single-tissue bias when comparing pQTLs to results of multi-tissue eQTLs.

    “This is the first study to address the single-tissue bias when comparing pQTLs to results of multi-tissue eQTLs.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    Compounds targeting proteins that are supported by genetic data are more likely to work than those without such support.

    “It is known that compounds targeting proteins that are supported by genetic data are more likely to work than those without such support42.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [42] Nelson, M. R. et al. The support of human genetic evidence for approved drug indications. Nat. Genet. 47, 856–860 (2015). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The percentages for coding variants were similar (Proportional-test: p-value > 0.05 in all three tissues), indicating that the enrichment of coding variants was not exclusively due to the bias of a single technology.

    “The percentages for coding variants were similar (Proportional-test: p-value > 0.05 in all three tissues), indicating that this enrichment was not exclusively due to the bias of a single technology.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The sample size of this study is similar to or larger than GTEx for brain-relevant tissues, and eQTL studies from GTEx do not see the same amount of coding-variant enrichment.

    “However, the sample size of our study is similar to or larger than GTEx for brain-relevant tissues, and eQTL studies from GTEx do not see the same amount of coding-variant enrichment5.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] Aguet, F. et al. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369, 1318–1330 (2020). source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Gene regulation may be region-specific, complicating overlap between QTL types.

    “The gene regulation may be region-specific, and thus complicates the overlapping between two QTL types.”
    Quote in the paperNot a substantial claim

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

    Interpretationnone given

    The small proportion of overlap between eQTL and pQTL data can also be caused by measurements of mRNA expression and protein level in different cohorts, sample sizes, or other technical details.

    “The small proportion of overlap between eQTL and pQTL data can also be caused by measurements of mRNA expression and protein level in different cohorts, sample sizes, or other technical details.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    Carvedilol has been reported to provide neuroprotection in animal models.

    “Carvedilol was identified from our drug repositioning analysis and has been reported to provide neuroprotection in animal models41.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41] Savitz, S. I. et al. The novel beta-blocker, carvedilol, provides neuroprotection in transient focal stroke. J. Cereb. Blood Flow. Metab. 20, 1197–1204 (2000). source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    For CSF, 965 out of 971 unique subjects have both genotype and proteomic data; for plasma, 633 out of 636; for brain, 450 out of 458.

    “For CSF, 965 out of 971 unique subjects have both genotype and proteomic data. For plasma, 633 out of 636 unique subjects have both genotype and proteomic data. For brain tissue, 450 out of 458 unique subjects have both genotype and proteomic data.”
    Quote in the paperNot a substantial claim

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

    Study design and conductnone given

    Data collection for the proteomics was randomized; all samples were randomly assigned to a profiling pool prior to proteomics measurement, considering age, sex, and disease status.

    “Data collection for the proteomics was randomized. All samples were randomly assigned to a profiling pool prior to proteomics measurement, considering age, sex, and disease status.”
    Quote in the paperNot a substantial claim

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

    Study design and conductown data

    After quality control, 931 plasma proteins were retained.

    “After step-5, 931 proteins that shared by < 10 samples were kept.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premiseown data

    The assay covers a dynamic range of 10^8 and measures all three major categories: secreted, membrane, and intracellular proteins.

    “The assay covers a dynamic range of 108, and measures all three major categories: secreted, membrane, and intracellular proteins (Table S34, Supplementary Fig 2, 3).”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites Table S34

      Not checked in the free run.

    • cites Supplementary Fig 2, 3

      Not checked in the free run.

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

    Study design and conductown data

    Sample sizes for the AD case-only analysis were 217 for CSF, 168 for plasma, and 248 for brain.

    “Sample size for this AD case (CA) only analysis was 217, 168, and 248 for CSF, plasma, and brain respectively.”
    Quote in the paperNot a substantial claim

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

    Study design and conductown data

    Sample sizes for the joint analysis were 835 for CSF, 529 for plasma, and 380 for brain.

    “Sample size for this joint analysis was 835, 529, and 380 for CSF, plasma, and brain respectively.”
    Quote in the paperNot a substantial claim

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

    Premisecitation

    The coverage of proteins on the platform has taken into account proteins known to be relevant to human disease, including neurodegenerative diseases and cardiovascular diseases.

    “The coverage of proteins on the platform has taken into account proteins known to be relevant to human disease, including neurodegenerative diseases55 and cardiovascular diseases56; thus it has been widely used for biomarker discovery.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [55] Sattlecker, M. et al. Alzheimer’s disease biomarker discovery using SOMAscan multiplexed protein technology. Alzheimers Dement. 10, 724–734 (2014). source

      Not checked in the free run.

    • cites [56] Williams, S. A. et al. Plasma protein patterns as comprehensive indicators of health. Nat. Med. 25, 1851–1857 (2019). source

      Not checked in the free run.

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

    Study design and conductcitation

    No statistical methods were used to predetermine sample sizes within each tissue type, but the sample sizes are much larger or comparable to those in previous publications.

    “No statistical methods were used to predetermine sample sizes within each tissue type, but our sample sizes are much larger or comparable to those in previous publications13,14,16.”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [13] Deming, Y. et al. Genetic studies of plasma analytes identify novel potential biomarkers for several complex traits. Sci. Rep. 6, 18092 (2016). source

      Not checked in the free run.

    • cites [14] Sasayama, D. et al. Genome-wide quantitative trait loci mapping of the human cerebrospinal fluid proteome. Hum. Mol. Genet. 26, 44–51 (2017).

      Not checked in the free run.

    • cites [16] Robins, C. et al. Genetic control of the human brain proteome. Preprint at bioRxiv https://doi.org/10.1101/816652 (2019). source

      Not checked in the free run.

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

    Premisecitation

    DrugBank is a manually curated database that maintains profiles for >15,000 drugs, including FDA-approved and experimental drugs.

    “This is a manually curated database that maintains profiles for >15,000 drugs (including FDA-approved and experimental drugs).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [73] Wishart, D. S. et al. DrugBank: a comprehensive resource for in silico drug discovery and exploration. Nucleic Acids Res. 34, D668–D672 (2006). source

      Not checked in the free run.

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

    Study design and conductown data

    After quality control, 1079 brain proteins were retained.

    “After step-5, 1079 proteins that shared by < 21 samples were kept.”
    Quote in the paperNot a substantial claim

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

    Study design and conductnone given

    Data collection for the genotype was not randomized or blocked.

    “Data collection for the genotype was not randomized or blocked.”
    Quote in the paperNot a substantial claim

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

    Study design and conductown data

    After quality control, 713 CSF proteins were retained.

    “After step-5, 713 proteins that shared by < 30 samples (shared by ~80% of the subject outliers) were kept.”
    Quote in the paperNot a substantial claim

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

    Study design and conductown data

    Sample sizes for the cognitively unimpaired (CO) only analysis were 614 for CSF, 357 for plasma, and 24 for brain.

    “Sample size for this cognitive unimpaired (CO) only analysis was 614, 357, and 24 for CSF, plasma, and brain respectively.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The pseudo-bulk derived cell-type specific eQTLs were underpowered compared with bulk-level pQTLs.

    “We used different priors because the pseudo-bulk derived cell-type specific eQTLs were underpowered compared with bulk-level pQTLs with p1 as 1×10−4, p2 as 1×10−2, and p12 as 1×10−3.”
    Quote in the paperSubstantial claim
46 of 46

  • The paper relies on this without support.

    Most GWAS identify multiple disease loci rather than functional variants or genes.

    “Most GWAS identify multiple disease loci rather than functional variants or genes, which makes it difficult to biologically interpret association results and to identify novel biomarkers and drug targets.”

  • The paper relies on this without support.

    eQTL mapping has not been able to fully identify the functional variants and genes driving GWAS signals.

    “However, eQTL mapping has not been able to fully identify the functional variants and genes driving GWAS signals.”

  • The authors state this as an assumption.

    Data distribution was assumed to be normal, but this was not formally tested.

    “For these tests, data distribution was assumed to be normal, but this was not formally tested.”
    “For these tests, data distribution was assumed to be normal, but this was not formally tested.”

  • The paper relies on this without support.

    Understanding the genetic architecture of protein levels is necessary to fully understand the genetic architecture of complex traits.

    “If we want to fully understand the genetic architecture of complex traits, we need to understand the genetic architecture of protein levels.”

  • The paper relies on this without support.

    Until now, multi-tissue pQTL mapping has been constrained by the limited availability of large-scale proteomic analyses in multiple tissues.

    “Until now, multi-tissue pQTL mapping has been constrained by the limited availability of large-scale proteomic analyses in multiple tissues.”
5 of 5

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?

  • Plasma proteogenomics shares little overlap with the brain.

    “Large studies have investigated genetic association of protein levels, but they have overwhelmingly focused on plasma11–15; previous work has shown that plasma proteogenomics shares little overlap with the brain10.”

    From Proteogenomic analysis of human cerebrospinal fluid identifies neurologically relevant regulation and implicates causal proteins for Alzheimer’s disease (2024) · cites this paper as [10]

    The Results section (‘Tissue-specificity investigation’; page not specified in our copy) says plasma shared more than 70% of brain pQTLs and describes plasma as informative for studying brain-related disorders. That is contrary to the claim of little overlap. (LLM, conf=0.90)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

  • Another recent CSF pQTL study also implicated CD33 for AD by MR analysis.

    “We note that another recent CSF pQTL study also implicated CD33 for AD by MR analysis (Yang et al, 2021), while several other proteins were uniquely implicated in this study versus in Yang et al (2021).”

    From The genetic regulation of protein expression in cerebrospinal fluid (2023) · cites this paper as Yang et al, 2021

    The Results section (“Mendelian randomization and drug repositioning”; page unavailable) says MR implicated CD33 in AD, and the study did analyze CSF pQTLs. However, it identifies the CD33–AD result specifically as a plasma-protein finding, not a CSF CD33 finding. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • Two pQTLs were common in both studies but associated with different proteins: rs76904798 was trans-acting and associated with GPNMB in this study and GRN in Yang et al (2021); rs12126142 was trans-acting in this study (PSME1) but cis-acting in Yang et al (2021) (IL6R).

    “We also identified two additional pQTLs that were common in both studies but associated with different proteins. One of these (rs76904798) was trans‐acting and was associated with GPNMB in our study and GRN in Yang et al (2021). The second pQTL (rs12126142) was trans‐acting in our study (PSME1) but cis‐acting in the Yang et al (2021) study (IL6R).”

    From The genetic regulation of protein expression in cerebrospinal fluid (2023) · cites this paper as Yang et al, 2021

    Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • Fifteen pQTL-protein pairs were replicated between this study and Yang et al (2021) with significant effects and similar direction; 12 were cis-acting and 3 were trans-acting.

    “Fifteen pQTL‐protein pairs were replicated (significant effects with similar direction of effect with respect to the risk allele) between the studies (see Fig EV4; Dataset EV7). Out of these 15, 12 pQTL‐protein pairs were cis‐acting and three were trans‐acting.”

    From The genetic regulation of protein expression in cerebrospinal fluid (2023) · cites this paper as Yang et al, 2021

    Split into 2 atoms — 1 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

    Experimental — accuracy not yet measured

  • cis-pQTL effect sizes correlate with proximity to the protein coding gene.

    “We also confirmed previous evidence10,13 that cis-pQTL effect sizes correlate with proximity to the protein coding gene (Fig. 2D; Supp. Fig. 3H).”

    From Proteogenomic analysis of human cerebrospinal fluid identifies neurologically relevant regulation and implicates causal proteins for Alzheimer’s disease (2024) · cites this paper as [10]

    The Results section on functional annotation reports that cis-pQTL association strength (−log10 P) decreased with distance from the transcription start site in all three tissues. Extended Data Fig. 10 likewise plots association significance against distance, but plots effect size (beta) against minor allele frequency instead. The copy therefore supports a proximity relationship for association significance, not specifically for effect size. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • PPMI replication sample count differs

    Minor

    The results give 131 PPMI samples, whereas the methods say 132 PPMI CSF samples passed both genomic and proteomic QC.

  • Conflicting relatedness thresholds

    Minor

    The genomic QC methods say one sample was removed from a pair at PI_HAT ≥ 0.5. The Extended Data Fig. 1 legend instead says retained donors had PI_HAT < 0.05, a tenfold stricter threshold.

  • Disease-stratified brain correlation contradicts the reported minimum

    Minor

    The results state that all disease-stratified effect-size correlations exceeded 0.98, but the methods report a correlation of 0.602 for the brain control-only analysis. The lower value qualifies the conclusion that the associations are not disease-specific.

3 of 3

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Yang, C., Farias, F. H. G., Ibanez, L., Suhy, A., Sadler, B., Fernandez, M. V., Wang, F., Bradley, J. L., Eiffert, B., Bahena, J. A., Budde, J. P., Li, Z., Dube, U., Sung, Y. J., Mihindukulasuriya, K. A., Morris, J. C., Fagan, A. M., Perrin, R. J., Benitez, B. A., Rhinn, H., Harari, O., & Cruchaga, C.
Affiliations
Washington University in St. Louis (US)
Corresponding author
Cruchaga, Carlos
ORCID
Venue
Nature Neuroscience
Publisher
Springer Science and Business Media LLC
Published
2021-07-08
Volume
24
Issue
9
Pages
1302-1312
ISSN
  • 1097-6256 (ISSN-L)
  • 1546-1726
DOI
10.1038/s41593-021-00886-6
PMC
PMC8521603
OpenAlex
W3181408683
Semantic Scholar
0d589c7cbb4d5c3ebc31faec1517408030723fe2
Cited by
290
References
93
Citations by year
2026: 42 · 2025: 75 · 2024: 75 · 2023: 55 · 2022: 34 · 2021: 8 · 2020: 1
FWCI
25.30
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Neuroscience (publishedVersion)
  • Research Square (acceptedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (acceptedVersion)
Funders
U.S. Department of Health & Human Services | NIH | National Institute on Aging (R01AG044546, RF1AG053303)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Mapping and Diversity in Plants and Animals
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
Keywords
protein quantitative trait loci, proteome, brain, cerebrospinal fluid, plasma, Alzheimer's disease, Parkinson's disease, stroke, neurological disorders, post-transcriptional regulation, tissue-specific gene regulation, Mendelian randomization, druggable targets
MeSH terms
Aged, Alzheimer Disease, Brain, Cerebrospinal Fluid, Female, Gene Expression Regulation, Humans, Male, Middle Aged, Plasma, Proteome, Quantitative Trait Loci, Proteomics, High-Throughput Screening Assays

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 2026-10-10.

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

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