OTTERS: A powerful TWAS framework leveraging summary-level reference data

Dai, Qile; Zhou, Geyu; Zhao, Hongyu; Võsa, Urmo; Franke, Lude; Battle, Alexis; Teumer, Alexander; Lehtimäki, Terho; Raitakari, Olli T.; Esko, Tõnu; Epstein, Michael P.; Yang, Jingjing

doi:10.1101/2022.03.30.486451

Cited by 7 publications

(17 citation statements)

References 85 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By applying OTTERS 11 to the summary-level pQTL reference data of three tissues (brain, CSF and plasma) 5 and recent large-scale GWAS summary data of AD dementia, we obtained PWAS p-values with pQTL weights estimated by five complimentary PRS methods. Moderate inflation was observed in the Quantile-Quantile (Q-Q) plots of these proteome-wide p-values in all three tissues ( Fig.…”

Section: Resultsmentioning

confidence: 99%

“…Whereas, OTTERS 11 estimates w from the summary-level pQTL reference data that are assumed to be generated based on the following single variant linear regression models with standardized genotype vectors x j for genetic variants j = 1, …, m : Summary-level pQTL reference data include the marginal least squared effect estimates , sample sizes, and linkage disequilibrium coefficients in the pQTL cohort (which can also be derived from an external reference panel with the same ancestry as the pQTL cohort). OTTERS employs five representative PRS methods, including the P-value Thresholding with linkage disequilibrium (LD) clumping ( P+T ) 17 with p-value thresholds of 0.05 and 0.001, frequentist LASSO regression ( lassosum ) 18,19 , nonparametric Bayesian Dirichlet process regression ( SDPR ) 20,21 , and Bayesian multiple linear regression with continuous shrinkage prior ( PRS-CS ) 22 .…”

Section: Methodsmentioning

confidence: 98%

“…We first applied OTTERS 11 to estimate pQTL weights from the recently released summary-level pQTL data of brain (n=380), CSF (n=835), and plasma (n=529) 12 . These summary-level pQTL data were generated by using proteomics data of individuals with AD and cognitively normal individuals of European ancestry profiled from an aptamer-based platform 25 .…”

Section: Methodsmentioning

confidence: 99%

See 2 more Smart Citations

Proteome-wide association studies using summary proteomic data identified 23 risk genes of Alzheimer’s disease

Hu,

Dai,

Epstein

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

Characterizing the genetic mechanisms underlying Alzheimer's disease (AD) dementia is crucial for developing new therapeutics. Proteome-wide association study (PWAS) integrating proteomics data with genome-wide association study (GWAS) summary data was shown as a powerful tool for detecting risk genes. The identified PWAS risk genes can be interpretated as having genetic effects mediated through the genetically regulated protein abundances. Existing PWAS analyses of AD often rely on the availability of individual-level proteomics and genetics data of a reference cohort. Leveraging summary-level protein quantitative trait loci (pQTL) reference data of multiple relevant tissues is expected to improve PWAS findings for studying AD. Here, we applied our recently developed OTTERS tool to conduct PWAS of AD dementia, by leveraging summary-level pQTL data of brain, cerebrospinal fluid (CSF), and plasma tissues, and multiple statistical methods. For each target protein, imputation models of the protein abundance with genetic predictors were trained from summary-level pQTL data, estimating a set of pQTL weights for considered genetic predictors. PWAS p-values were obtained by integrating GWAS summary data of AD dementia with estimated pQTL weights. PWAS p-values from multiple statistical methods were combined by the aggregated Cauchy association test to yield one omnibus PWAS p-value for the target protein. We identified significant PWAS risk genes through omnibus PWAS p-values and analyzed their protein-protein interactions using STRING. Their potential causal effects were assessed by the probabilistic Mendelian randomization (PMR-Egger). As a result, we identified a total of 23 significant PWAS risk genes for AD dementia in brain, CSF, and plasma tissues, including 7 novel findings. We showed that 15 of these risk genes were interconnected within a protein-protein interaction network involving the well-known AD risk gene of APOE and 5 novel findings, and enriched in immune functions and lipids pathways including positive regulation of immune system process, positive regulation of macrophage proliferation, humoral immune response, and high-density lipoprotein particle clearance. Existing biological evidence was found to relate our novel findings with AD. We validated the mediated causal effects of 14 risk genes (60.8%). In conclusion, we identified both known and novel PWAS risk genes, providing novel insights into the genetic mechanisms in brain, CSF, and plasma tissues, and targeted therapeutics development of AD dementia. Our study also demonstrated the effectiveness of integrating public available summary-level pQTL data with GWAS summary data for mapping risk genes of complex human diseases.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 98%

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Proteome-wide association studies using summary proteomic data identified 23 risk genes of Alzheimer’s disease

Hu,

Dai,

Epstein

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…A recent preprint describes a similar approach to ours for the integration of eQTL summary statistics in a TWAS framework, called OTTERS (30). Our findings are congruent, with summary statistic methods applied to eQTLGen providing novel opportunities and associations.…”

Section: Discussionmentioning

confidence: 99%

“…ACAT-O is an aggregated Cauchy association omnibus test, which combines p -values without needing to specify the correlation between the models for each gene (23). This method was recently proposed as an omnibus test for TWAS models (30). ACAT-O is robust when combining p -values corresponding to different directions of effect, so can be applied to two-sided p -values from TWAS.…”

Section: Methodsmentioning

confidence: 99%

Polygenic Prediction of Molecular Traits using Large-Scale Meta-analysis Summary Statistics

Pain

Gerring

Derks

et al. 2022

Preprint

View full text Add to dashboard Cite

Introduction: Transcriptome-wide association study (TWAS) integrates expression quantitative trait loci (eQTL) data with genome-wide association study (GWAS) results to infer differential expression. TWAS uses multi-variant models trained using individual-level genotype-expression datasets, but methodological development is required for TWAS to utilise larger eQTL summary statistics. Methods: TWAS models predicting gene expression were derived using blood-based eQTL summary statistics from eQTLGen, the Young Finns Study (YFS), and MetaBrain. Summary statistic polygenic scoring methods were used to derive TWAS models, evaluating their predictive utility in GTEx v8. We investigated gene inclusion criteria and omnibus tests for aggregating TWAS associations for a given gene. We performed a schizophrenia TWAS using summary statistic-based TWAS models, comparing results to existing resources and methods. Results: TWAS models derived using eQTL summary statistics performed comparably to models derived using individual-level data. Multi-variant TWAS models significantly improved prediction over single variant models for 8.6% of genes. TWAS models derived using eQTLGen summary statistics significantly improved prediction over models derived using a smaller individual-level dataset. The eQTLGen-based schizophrenia TWAS, using the ACAT omnibus test to aggregate associations for each gene, identified novel significant and colocalised associations compared to summary-based mendelian randomisation (SMR) and SMR-multi. Conclusions: Using multi-variant TWAS models and larger eQTL summary statistic datasets can improve power to detect differential expression associations. We provide TWAS models based on eQTLGen and MetaBrain summary statistics, and software to easily derive and apply summary statistic-based TWAS models based on eQTL and other molecular QTL datasets released in the future.

show abstract

Omnibus proteome-wide association study (PWAS-O) identified 43 risk genes for Alzheimer’s disease dementia

Parrish

Buchman

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Proteome-wide association study (PWAS) integrating proteomics data with GWAS summary data is a powerful tool to identify risk genes for complex diseases. These genes can inform disease mechanisms with genetic effects mediated through protein abundance. We propose a novel omnibus method to further improve PWAS power by modeling unknown genetic architectures with multiple statistical models. We applied TIGAR, PrediXcan, and FUSION to train protein abundance imputation models for 8,430 proteins from dorsolateral prefrontal cortex with whole genome sequencing data (n=355). Next, the trained models were integrated with GWAS summary data of Alzheimer's disease (AD) dementia (n=~70K) to conduct PWAS. Last, we employed the Aggregated Cauchy Association Test to obtain omnibus PWAS (PWAS-O) p-values from these three tools. PWAS-O identified 43 risk genes of AD dementia that were interconnected into a protein-protein interaction network including TOMM40, APOC1, and APOC2. It detected 5 novel genes missed by GWAS and TWAS. PWAS-O can be applied to characterize underlying genetic mechanisms for diverse diseases.

show abstract

OTTERS: A powerful TWAS framework leveraging summary-level reference data

Cited by 7 publications

References 85 publications

Proteome-wide association studies using summary proteomic data identified 23 risk genes of Alzheimer’s disease

Proteome-wide association studies using summary proteomic data identified 23 risk genes of Alzheimer’s disease

Polygenic Prediction of Molecular Traits using Large-Scale Meta-analysis Summary Statistics

Omnibus proteome-wide association study (PWAS-O) identified 43 risk genes for Alzheimer’s disease dementia

Contact Info

Product

Resources

About