Identifying and correcting for misspecifications in GWAS summary statistics and polygenic scores

Privé, Florian; Bj, Vilhjalmsson; Tsh, Mak

doi:10.1101/2021.03.29.437510

Cited by 15 publications

(11 citation statements)

References 65 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yet, there are dramatic overall trends as seen in Figure A5. In general, even though the quartiles contain approximately 3 times as many SNPs as the HapMapIII SNP set, when those SNPs are poorly imputed the performance of Vilma suffers substantially, which has also been ovserved in [41]. Indeed for the lowest and second lowest quartiles, we see huge drops in performance when looking across traits and target cohorts (p 10 −16 in both cases with median drops in r of 0.08 and 0.02 respectively).…”

Section: A3 Using Poorly Imputed Variants Can Degrade Performancementioning

confidence: 88%

A flexible modeling and inference framework for estimating variant effect sizes from GWAS summary statistics

Spence

Sinnott-Armstrong

Assimes

et al. 2022

Preprint

View full text Add to dashboard Cite

Genome-wide association studies (GWAS) have highlighted that almost any trait is affected by many variants of relatively small effect. On one hand this presents a challenge for inferring the effect of any single variant as the signal-to-noise ratio is high for variants of small effect. This challenge is compounded when combining information across many variants in polygenic scores for predicting trait values. On the other hand, the large number of contributing variants provides an opportunity to learn about the average behavior of variants encoded in the distribution of variant effect sizes. Many approaches have looked at aspects of this problem, but no method has unified the inference of the effects of individual variants with the inference of the distribution of effect sizes while requiring only GWAS summary statistics and properly accounting for linkage disequilibrium between variants. Here we present a flexible, unifying framework that combines information across variants to infer a distribution of effect sizes and uses this distribution to improve the estimation of the effects of individual variants. We also develop a variational inference (VI) scheme to perform efficient inference under this framework. We show this framework is useful by constructing polygenic scores (PGSs) that outperform the state-of-the-art. Our modeling framework easily extends to jointly inferring effect sizes across multiple cohorts, where we show that building PGSs using additional cohorts of differing ancestries improves predictive accuracy and portability. We also investigate the inferred distributions of effect sizes across many traits and find that these distributions have effect sizes ranging over multiple orders of magnitude, in contrast to the assumptions implicit in many commonly-used statistical genetics methods.

show abstract

Section: A3 Using Poorly Imputed Variants Can Degrade Performancementioning

confidence: 88%

A flexible modeling and inference framework for estimating variant effect sizes from GWAS summary statistics

Spence

Sinnott-Armstrong

Assimes

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This is in contrast to most of the statistical methods used above where the inclusion of biomarkers relied mostly on previous research on the biomarkers themselves. In addition to feature selection, this sparse approach has been chosen because of previous success with SNP based prediction [ 11 , 54 , 55 , 56 , 57 , 58 , 59 ] and because it has been shown to be among the best ML predictors for genetics and is often a good all around method [ 56 , 60 ].…”

Section: Introductionmentioning

confidence: 99%

Machine Learning Prediction of Biomarkers from SNPs and of Disease Risk from Biomarkers in the UK Biobank

Widén

Raben

Lello

et al. 2021

Genes

View full text Add to dashboard Cite

We use UK Biobank data to train predictors for 65 blood and urine markers such as HDL, LDL, lipoprotein A, glycated haemoglobin, etc. from SNP genotype. For example, our Polygenic Score (PGS) predictor correlates ∼0.76 with lipoprotein A level, which is highly heritable and an independent risk factor for heart disease. This may be the most accurate genomic prediction of a quantitative trait that has yet been produced (specifically, for European ancestry groups). We also train predictors of common disease risk using blood and urine biomarkers alone (no DNA information); we call these predictors biomarker risk scores, BMRS. Individuals who are at high risk (e.g., odds ratio of >5× population average) can be identified for conditions such as coronary artery disease (AUC∼0.75), diabetes (AUC∼0.95), hypertension, liver and kidney problems, and cancer using biomarkers alone. Our atherosclerotic cardiovascular disease (ASCVD) predictor uses ∼10 biomarkers and performs in UKB evaluation as well as or better than the American College of Cardiology ASCVD Risk Estimator, which uses quite different inputs (age, diagnostic history, BMI, smoking status, statin usage, etc.). We compare polygenic risk scores (risk conditional on genotype: PRS) for common diseases to the risk predictors which result from the concatenation of learned functions BMRS and PGS, i.e., applying the BMRS predictors to the PGS output.

show abstract

“…In other cases, the GWAS summary statistics may be derived from a meta-analysis that combines data from a number of different studies. These settings may present potential mismatches and heterogeneities between of LD reference panel and GWAS summary statistics and are thus challenging to model, often leading to substantial loss in predictive power [29, 30, 68, 69].…”

Section: Resultsmentioning

confidence: 99%

“…Summary statistics-based PRS methods can be sensitive to heterogeneities between GWAS summary statistics and the LD reference panel [8,29,68]. For some Bayesian methods, this mismatch can result in unpredictable behavior, with the posterior mean for the effect sizes ex-ploding in magnitude and the estimated SNP-heritability exceeding 1 in some circumstances 1 .…”

Section: S16 a Heuristic Test Of Mismatch Between Gwas Summary Statis...mentioning

confidence: 99%

Fast and Accurate Bayesian Polygenic Risk Modeling with Variational Inference

Zabad

Gravel

2022

Preprint

View full text Add to dashboard Cite

The recent proliferation of large scale genome-wide association studies (GWASs) has motivated the development of statistical methods for phenotype prediction using single nucleotide polymorphism (SNP) array data. These polygenic risk score (PRS) methods formulate the task of polygenic prediction in terms of a multiple linear regression framework, where the goal is to infer the joint effect sizes of all genetic variants on the trait. Among the subset of PRS methods that operate on GWAS summary statistics, sparse Bayesian methods have shown competitive predictive ability. However, existing Bayesian approaches employ Markov Chain Monte Carlo (MCMC) algorithms for posterior inference, which are computationally inefficient and do not scale favorably with the number of SNPs included in the analysis. Here, we introduce Variational Inference of Polygenic Risk Scores (VIPRS), a Bayesian summary statistics-based PRS method that utilizes Variational Inference (VI) techniques to efficiently approximate the posterior distribution for the effect sizes. Our experiments with genome-wide simulations and real phenotypes from the UK Biobank (UKB) dataset demonstrated that variational approximations to the posterior are competitively accurate and highly efficient. When compared to state-of-the-art PRS methods, VIPRS consistently achieves the best or second best predictive accuracy in our analyses of 18 simulation configurations as well as 12 real phenotypes measured among the UKB participants of “White British” background. This performance advantage was higher among individuals from other ethnic groups, with an increase in R2 of up to 1.7-fold among participants of Nigerian ancestry for Low-Density Lipoprotein (LDL) cholesterol. Furthermore, given its computational efficiency, we applied VIPRS to a dataset of up to 10 million genetic markers, an order of magnitude greater than the standard HapMap3 subset used to train existing PRS methods. Modeling this expanded set of variants conferred modest improvements in prediction accuracy for a number of highly polygenic traits, such as standing height.

show abstract

Identifying and correcting for misspecifications in GWAS summary statistics and polygenic scores

Cited by 15 publications

References 65 publications

A flexible modeling and inference framework for estimating variant effect sizes from GWAS summary statistics

A flexible modeling and inference framework for estimating variant effect sizes from GWAS summary statistics

Machine Learning Prediction of Biomarkers from SNPs and of Disease Risk from Biomarkers in the UK Biobank

Fast and Accurate Bayesian Polygenic Risk Modeling with Variational Inference

Contact Info

Product

Resources

About