LDpred2: better, faster, stronger

Abstract Motivation Polygenic scores have become a central tool in human genetics research. LDpred is a popular method for deriving polygenic scores based on summary statistics and a matrix of correlation between genetic variants. However, LDpred has limitations that may reduce its predictive performance. Results Here, we present LDpred2, a new version of LDpred that addresses these issues. We also provide two new options in LDpred2: a ‘sparse’ option that can learn effects that are exactly 0, and an ‘auto’ option that directly learns the two LDpred parameters from data. We benchmark predictive performance of LDpred2 against the previous version on simulated and real data, demonstrating substantial improvements in robustness and predictive accuracy compared to LDpred1. We then show that LDpred2 also outperforms other polygenic score methods recently developed, with a mean AUC over the 8 real traits analyzed here of 65.1%, compared to 63.8% for lassosum, 62.9% for PRS-CS and 61.5% for SBayesR. Note that LDpred2 provides more accurate polygenic scores when run genome-wide, instead of per chromosome. Availability and implementation LDpred2 is implemented in R package bigsnpr. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

LDpred2: better, faster, stronger

10.1101/2020.04.28.066720 ◽

2020 ◽

Cited By ~ 3

Author(s):

Florian Privé ◽

Julyan Arbel ◽

Bjarni J. Vilhjálmsson

Keyword(s):

Human Genetics ◽

Predictive Accuracy ◽

Predictive Performance ◽

Real Data ◽

R Package ◽

Summary Statistics ◽

Genetics Research ◽

Genome Wide ◽

Polygenic Scores ◽

Central Tool

AbstractPolygenic scores have become a central tool in human genetics research. LDpred is a popular method for deriving polygenic scores based on summary statistics and a matrix of correlation between genetic variants. However, LDpred has limitations that may reduce its predictive performance. Here we present LDpred2, a new version of LDpred that addresses these issues. We also provide two new options in LDpred2: a “sparse” option that can learn effects that are exactly 0, and an “auto” option that directly learns the two LDpred parameters from data. We benchmark predictive performance of LDpred2 against the previous version on simulated and real data, demonstrating substantial improvements in robustness and predictive accuracy compared to LDpred1. We then show that LDpred2 also outperforms other polygenic score methods recently developed, with a mean AUC over the 8 real traits analyzed here of 65.1%, compared to 63.8% for lassosum, 62.9% for PRS-CS and 61.5% for SBayesR. Note that, in contrast to what was recommended in the first version of this paper, we now recommend to run LDpred2 genome-wide instead of per chromosome. LDpred2 is implemented in R package bigsnpr.

Download Full-text

Interaction screening by Kendall’s partial correlation for ultrahigh-dimensional data with survival trait

Bioinformatics ◽

10.1093/bioinformatics/btaa017 ◽

2020 ◽

Vol 36 (9) ◽

pp. 2763-2769

Author(s):

Jie-Huei Wang ◽

Yi-Hau Chen

Keyword(s):

Partial Correlation ◽

Association Studies ◽

B Cell Lymphoma ◽

Real Data ◽

R Package ◽

Supplementary Information ◽

Genome Wide Association Studies ◽

Genome Wide ◽

Interaction Screening ◽

Relationship Of

Abstract Motivation In gene expression and genome-wide association studies, the identification of interaction effects is an important and challenging issue owing to its ultrahigh-dimensional nature. In particular, contaminated data and right-censored survival outcome make the associated feature screening even challenging. Results In this article, we propose an inverse probability-of-censoring weighted Kendall’s tau statistic to measure association of a survival trait with biomarkers, as well as a Kendall’s partial correlation statistic to measure the relationship of a survival trait with an interaction variable conditional on the main effects. The Kendall’s partial correlation is then used to conduct interaction screening. Simulation studies under various scenarios are performed to compare the performance of our proposal with some commonly available methods. In the real data application, we utilize our proposed method to identify epistasis associated with the clinical survival outcomes of non-small-cell lung cancer, diffuse large B-cell lymphoma and lung adenocarcinoma patients. Both simulation and real data studies demonstrate that our method performs well and outperforms existing methods in identifying main and interaction biomarkers. Availability and implementation R-package ‘IPCWK’ is available to implement this method, together with a reference manual describing how to perform the ‘IPCWK’ package. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

Detection of differentially methylated CpG sites between tumor samples with uneven tumor purities

Bioinformatics ◽

10.1093/bioinformatics/btz885 ◽

2019 ◽

Vol 36 (7) ◽

pp. 2017-2024

Author(s):

Weiwei Zhang ◽

Ziyi Li ◽

Nana Wei ◽

Hua-Jun Wu ◽

Xiaoqi Zheng

Keyword(s):

Real Data ◽

R Package ◽

Differential Methylation ◽

Least Square ◽

Epigenetic Mechanism ◽

Supplementary Information ◽

Cpg Sites ◽

Tumor Purity ◽

Different Sources ◽

Normal Controls

Abstract Motivation Inference of differentially methylated (DM) CpG sites between two groups of tumor samples with different geno- or pheno-types is a critical step to uncover the epigenetic mechanism of tumorigenesis, and identify biomarkers for cancer subtyping. However, as a major source of confounding factor, uneven distributions of tumor purity between two groups of tumor samples will lead to biased discovery of DM sites if not properly accounted for. Results We here propose InfiniumDM, a generalized least square model to adjust tumor purity effect for differential methylation analysis. Our method is applicable to a variety of experimental designs including with or without normal controls, different sources of normal tissue contaminations. We compared our method with conventional methods including minfi, limma and limma corrected by tumor purity using simulated datasets. Our method shows significantly better performance at different levels of differential methylation thresholds, sample sizes, mean purity deviations and so on. We also applied the proposed method to breast cancer samples from TCGA database to further evaluate its performance. Overall, both simulation and real data analyses demonstrate favorable performance over existing methods serving similar purpose. Availability and implementation InfiniumDM is a part of R package InfiniumPurify, which is freely available from GitHub (https://github.com/Xiaoqizheng/InfiniumPurify). Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

bGWAS: an R package to perform Bayesian genome wide association studies

Bioinformatics ◽

10.1093/bioinformatics/btaa549 ◽

2020 ◽

Vol 36 (15) ◽

pp. 4374-4376

Author(s):

Ninon Mounier ◽

Zoltán Kutalik

Keyword(s):

Mendelian Randomization ◽

Causal Effect ◽

Association Studies ◽

R Package ◽

Genome Wide Association ◽

Supplementary Information ◽

Genome Wide Association Studies ◽

Biological Mechanisms ◽

Genome Wide ◽

Related Risk

Abstract Summary Increasing sample size is not the only strategy to improve discovery in Genome Wide Association Studies (GWASs) and we propose here an approach that leverages published studies of related traits to improve inference. Our Bayesian GWAS method derives informative prior effects by leveraging GWASs of related risk factors and their causal effect estimates on the focal trait using multivariable Mendelian randomization. These prior effects are combined with the observed effects to yield Bayes Factors, posterior and direct effects. The approach not only increases power, but also has the potential to dissect direct and indirect biological mechanisms. Availability and implementation bGWAS package is freely available under a GPL-2 License, and can be accessed, alongside with user guides and tutorials, from https://github.com/n-mounier/bGWAS. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

powmic: an R package for power assessment in microbiome case–control studies

Bioinformatics ◽

10.1093/bioinformatics/btaa197 ◽

2020 ◽

Vol 36 (11) ◽

pp. 3563-3565

Author(s):

Li Chen

Keyword(s):

Power Analysis ◽

Real Data ◽

Analytical Form ◽

R Package ◽

Case Control ◽

Supplementary Information ◽

Metagenomic Sequencing ◽

Case Control Studies ◽

Simulation Based ◽

Over Dispersion

Abstract Summary Power analysis is essential to decide the sample size of metagenomic sequencing experiments in a case–control study for identifying differentially abundant (DA) microbes. However, the complexity of microbial data characteristics, such as excessive zeros, over-dispersion, compositionality, intrinsically microbial correlations and variable sequencing depths, makes the power analysis particularly challenging because the analytical form is usually unavailable. Here, we develop a simulation-based power assessment strategy and R package powmic, which considers the complexity of microbial data characteristics. A real data example demonstrates the usage of powmic. Availability and implementation powmic R package and online tutorial are available at https://github.com/lichen-lab/powmic. Contact [email protected] Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

Multi-SNP mediation intersection-union test

Bioinformatics ◽

10.1093/bioinformatics/btz285 ◽

2019 ◽

Vol 35 (22) ◽

pp. 4724-4729 ◽

Cited By ~ 4

Author(s):

Wujuan Zhong ◽

Cassandra N Spracklen ◽

Karen L Mohlke ◽

Xiaojing Zheng ◽

Jason Fine ◽

...

Keyword(s):

Association Studies ◽

R Package ◽

Alternative Methods ◽

Supplementary Information ◽

Genome Wide Association Studies ◽

Mediation Effects ◽

Coding Regions ◽

Genome Wide ◽

Plasma Adiponectin Level ◽

Intersection Union Test

Abstract Summary Tens of thousands of reproducibly identified GWAS (Genome-Wide Association Studies) variants, with the vast majority falling in non-coding regions resulting in no eventual protein products, call urgently for mechanistic interpretations. Although numerous methods exist, there are few, if any methods, for simultaneously testing the mediation effects of multiple correlated SNPs via some mediator (e.g. the expression of a gene in the neighborhood) on phenotypic outcome. We propose multi-SNP mediation intersection-union test (SMUT) to fill in this methodological gap. Our extensive simulations demonstrate the validity of SMUT as well as substantial, up to 92%, power gains over alternative methods. In addition, SMUT confirmed known mediators in a real dataset of Finns for plasma adiponectin level, which were missed by many alternative methods. We believe SMUT will become a useful tool to generate mechanistic hypotheses underlying GWAS variants, facilitating functional follow-up. Availability and implementation The R package SMUT is publicly available from CRAN at https://CRAN.R-project.org/package=SMUT. Supplementary information Supplementary data are available at Bioinformatics online.

Download Full-text

Polygenic score accuracy in ancient samples: quantifying the effects of allelic turnover

10.1101/2021.09.21.461259 ◽

2021 ◽

Author(s):

Maryn O. Carlson ◽

Daniel P. Rice ◽

Jeremy J. Berg ◽

Matthias Steinrücken

Keyword(s):

Human Genetics ◽

Transition Density ◽

Complex Trait ◽

Weak Selection ◽

Trait Variation ◽

Polygenic Score ◽

Error Metrics ◽

Genome Wide ◽

Polygenic Scores ◽

Evolutionary Context

AbstractPolygenic scores link the genotypes of ancient individuals to their phenotypes, which are often unobservable, offering a tantalizing opportunity to reconstruct complex trait evolution. In practice, however, interpretation of ancient polygenic scores is subject to numerous assumptions. For one, the genome-wide association (GWA) studies from which polygenic scores are derived, can only estimate effect sizes for loci segregating in contemporary populations. Therefore, a GWA study may not correctly identify all loci relevant to trait variation in the ancient population. In addition, the frequencies of trait-associated loci may have changed in the intervening years. Here, we devise a theoretical framework to quantify the effect of this allelic turnover on the statistical properties of polygenic scores as functions of population genetic dynamics, trait architecture, power to detect significant loci, and the age of the ancient sample. We model the allele frequencies of loci underlying trait variation using the Wright-Fisher diffusion, and employ the spectral representation of its transition density to find analytical expressions for several error metrics, including the correlation between an ancient individual’s polygenic score and true phenotype, referred to as polygenic score accuracy. Our theory also applies to a two-population scenario and demonstrates that allelic turnover alone may explain a substantial percentage of the reduced accuracy observed in cross-population predictions, akin to those performed in human genetics. Finally, we use simulations to explore the effects of recent directional selection, a bias-inducing process, on the statistics of interest. We find that even in the presence of bias, weak selection induces minimal deviations from our neutral expectations for the decay of polygenic score accuracy. By quantifying the limitations of polygenic scores in an explicit evolutionary context, our work lays the foundation for the development of more sophisticated statistical procedures to analyze both temporally and geographically resolved polygenic scores.

Download Full-text

Making the most of Clumping and Thresholding for polygenic scores

10.1101/653204 ◽

2019 ◽

Cited By ~ 1

Author(s):

Florian Privé ◽

Bjarni J. Vilhjálmsson ◽

Hugues Aschard ◽

Michael G.B. Blum

Keyword(s):

Predictive Ability ◽

Predictive Performance ◽

Real Data ◽

Penalized Regression ◽

P Value ◽

Single Node ◽

Depression Status ◽

Polygenic Scores ◽

Optimal Linear ◽

The Uk

AbstractPolygenic prediction has the potential to contribute to precision medicine. Clumping and Thresh-olding (C+T) is a widely used method to derive polygenic scores. When using C+T, it is common to test several p-value thresholds to maximize predictive ability of the derived polygenic scores. Along with this p-value threshold, we propose to tune three other hyper-parameters for C+T. We implement an efficient way to derive thousands of different C+T polygenic scores corresponding to a grid over four hyper-parameters. For example, it takes a few hours to derive 123,200 different C+T scores for 300K individuals and 1M variants on a single node with 16 cores.We find that optimizing over these four hyper-parameters improves the predictive performance of C+T in both simulations and real data applications as compared to tuning only the p-value threshold. A particularly large increase can be noted when predicting depression status, from an AUC of 0.557 (95% CI: [0.544-0.569]) when tuning only the p-value threshold in C+T to an AUC of 0.592 (95% CI: [0.580-0.604]) when tuning all four hyper-parameters we propose for C+T.We further propose Stacked Clumping and Thresholding (SCT), a polygenic score that results from stacking all derived C+T scores. Instead of choosing one set of hyper-parameters that maximizes prediction in some training set, SCT learns an optimal linear combination of all C+T scores by using an efficient penalized regression. We apply SCT to 8 different case-control diseases in the UK biobank data and find that SCT substantially improves prediction accuracy with an average AUC increase of 0.035 over standard C+T.

Download Full-text

The power of pathway-based polygenic risk scores

10.21203/rs.3.rs-643696/v1 ◽

2021 ◽

Author(s):

Paul O’Reilly ◽

Shing Choi ◽

Judit Garcia-Gonzalez ◽

Yunfeng Ruan ◽

Hei Man Wu ◽

...

Keyword(s):

Complex Disease ◽

Predictive Accuracy ◽

Risk Scores ◽

Distinct Advantage ◽

Genetic Profile ◽

Genetic Liability ◽

Polygenic Risk ◽

Individual Level ◽

Genome Wide ◽

Polygenic Scores

Abstract Polygenic risk scores (PRSs) have been among the leading advances in biomedicine in recent years. As a proxy of genetic liability, PRSs are utilised across multiple fields and applications. While numerous statistical and machine learning methods have been developed to optimise their predictive accuracy, all of these distil genetic liability to a single number based on aggregation of an individual’s genome-wide alleles. This results in a key loss of information about an individual’s genetic profile, which could be critical given the functional sub-structure of the genome and the heterogeneity of complex disease. Here we evaluate the performance of pathway-based PRSs, in which polygenic scores are calculated across genomic pathways for each individual, and we introduce a software, PRSet, for computing and analysing pathway PRSs. We find that pathway PRSs have similar power for evaluating pathway enrichment of GWAS signal as the leading methods, with the distinct advantage of providing estimates of pathway genetic liability at the individual-level. Exemplifying their utility, we demonstrate that pathway PRSs can stratify diseases into subtypes in the UK Biobank with substantially greater power than genome-wide PRSs. Compared to genome-wide PRSs, we expect pathway-based PRSs to offer greater insights into the heterogeneity of complex disease and treatment response, generate more biologically tractable therapeutic targets, and provide a more powerful path to precision medicine.

Download Full-text

gwasurvivr: an R package for genome wide survival analysis

10.1101/326033 ◽

2018 ◽

Author(s):

Abbas A Rizvi ◽

Ezgi Karaesmen ◽

Martin Morgan ◽

Leah Preus ◽

Junke Wang ◽

...

Keyword(s):

Survival Analysis ◽

Cox Model ◽

R Package ◽

Supplementary Information ◽

Parameter Estimates ◽

Survival Analyses ◽

Link Type ◽

Genome Wide ◽

Size Number ◽

Simple Interface

ABSTRACTSummaryTo address the limited software options for performing survival analyses with millions of SNPs, we developed gwasurvivr, an R/Bioconductor package with a simple interface for conducting genome wide survival analyses using VCF (outputted from Michigan or Sanger imputation servers), IMPUTE2 or PLINK files. To decrease the number of iterations needed for convergence when optimizing the parameter estimates in the Cox model we modified the R package survival; covariates in the model are first fit without the SNP, and those parameter estimates are used as initial points. We benchmarked gwasurvivr with other software capable of conducting genome wide survival analysis (genipe, SurvivalGWAS_SV, and GWASTools). gwasurvivr is significantly faster and shows better scalability as sample size, number of SNPs and number of covariates increases.Availability and implementationgwasurvivr, including source code, documentation, and vignette are available at: http://bioconductor.org/packages/gwasurvivrContactAbbas Rizvi, [email protected]; Lara E Sucheston-Campbell, [email protected] information: Supplementary data are available at https://github.com/suchestoncampbelllab/gwasurvivr_manuscript

Download Full-text