Polygenic risk scores have shown great promise in predicting complex disease risk and will become more accurate as training sample sizes increase. The standard approach for calculating risk scores involves linkage disequilibrium (LD)-based marker pruning and applying a p value threshold to association statistics, but this discards information and can reduce predictive accuracy. We introduce LDpred, a method that infers the posterior mean effect size of each marker by using a prior on effect sizes and LD information from an external reference panel. Theory and simulations show that LDpred outperforms the approach of pruning followed by thresholding, particularly at large sample sizes. Accordingly, predicted R(2) increased from 20.1% to 25.3% in a large schizophrenia dataset and from 9.8% to 12.0% in a large multiple sclerosis dataset. A similar relative improvement in accuracy was observed for three additional large disease datasets and for non-European schizophrenia samples. The advantage of LDpred over existing methods will grow as sample sizes increase.

/pdf/modeling-linkage-disequilibrium-increases-accuracy-of-197j1wrd9y.pdf

Modeling Linkage Disequilibrium Increases Accuracy of Polygenic Risk Scores

Mixed linear models are emerging as a method of choice for conducting genetic association studies in humans and other organisms. The advantages of the mixed-linear-model association (MLMA) method include the prevention of false positive associations due to population or relatedness structure and an increase in power obtained through the application of a correction that is specific to this structure. An underappreciated point is that MLMA can also increase power in studies without sample structure by implicitly conditioning on associated loci other than the candidate locus. Numerous variations on the standard MLMA approach have recently been published, with a focus on reducing computational cost. These advances provide researchers applying MLMA methods with many options to choose from, but we caution that MLMA methods are still subject to potential pitfalls. Here we describe and quantify the advantages and pitfalls of MLMA methods as a function of study design and provide recommendations for the application of these methods in practical settings.

/pdf/advantages-and-pitfalls-in-the-application-of-mixed-model-3dn5v7uflg.pdf

Advantages and pitfalls in the application of mixed-model association methods

Both linear mixed models (LMMs) and sparse regression models are widely used in genetics applications, including, recently, polygenic modeling in genome-wide association studies. These two approaches make very different assumptions, so are expected to perform well in different situations. However, in practice, for a given dataset one typically does not know which assumptions will be more accurate. Motivated by this, we consider a hybrid of the two, which we refer to as a “Bayesian sparse linear mixed model” (BSLMM) that includes both these models as special cases. We address several key computational and statistical issues that arise when applying BSLMM, including appropriate prior specification for the hyper-parameters and a novel Markov chain Monte Carlo algorithm for posterior inference. We apply BSLMM and compare it with other methods for two polygenic modeling applications: estimating the proportion of variance in phenotypes explained (PVE) by available genotypes, and phenotype (or breeding value) prediction. For PVE estimation, we demonstrate that BSLMM combines the advantages of both standard LMMs and sparse regression modeling. For phenotype prediction it considerably outperforms either of the other two methods, as well as several other large-scale regression methods previously suggested for this problem. Software implementing our method is freely available from http://stephenslab.uchicago.edu/software.html.

/pdf/polygenic-modeling-with-bayesian-sparse-linear-mixed-models-2xqyvb27sa.pdf

Polygenic modeling with bayesian sparse linear mixed models.

Polygenic risk scores (PRS) have shown promise in predicting human complex traits and diseases. Here, we present PRS-CS, a polygenic prediction method that infers posterior effect sizes of single nucleotide polymorphisms (SNPs) using genome-wide association summary statistics and an external linkage disequilibrium (LD) reference panel. PRS-CS utilizes a high-dimensional Bayesian regression framework, and is distinct from previous work by placing a continuous shrinkage (CS) prior on SNP effect sizes, which is robust to varying genetic architectures, provides substantial computational advantages, and enables multivariate modeling of local LD patterns. Simulation studies using data from the UK Biobank show that PRS-CS outperforms existing methods across a wide range of genetic architectures, especially when the training sample size is large. We apply PRS-CS to predict six common complex diseases and six quantitative traits in the Partners HealthCare Biobank, and further demonstrate the improvement of PRS-CS in prediction accuracy over alternative methods.

/pdf/polygenic-prediction-via-bayesian-regression-and-continuous-3mjihzxubo.pdf

Polygenic prediction via Bayesian regression and continuous shrinkage priors.

The success of genome-wide association studies (GWASs) has led to increasing interest in making predictions of complex trait phenotypes, including disease, from genotype data. Rigorous assessment of the value of predictors is crucial before implementation. Here we discuss some of the limitations and pitfalls of prediction analysis and show how naive implementations can lead to severe bias and misinterpretation of results.

/pdf/pitfalls-of-predicting-complex-traits-from-snps-3mppaxq4eq.pdf

Pitfalls of predicting complex traits from SNPs

Improving accuracy of genomic predictions within and between dairy cattle breeds with imputed high-density single nucleotide polymorphism panels

Many quantitative trait loci (QTL) are located in non-coding genomic regions. Therefore, QTL are assumed to affect gene regulation. Gene expression and RNA splicing are primary steps of transcription so QTL changing gene expression (eQTL) or RNA splicing (sQTL) are expected to significantly contribute to phenotypic variations. Here, we quantify the contribution of eQTL and sQTL detected from 16 tissues (N~5,000) to 37 complex traits of ~120k cattle. Using Bayesian methods, we show that including more regulatory variants in the model explains larger proportions of heritability. Across traits, cis and trans eQTL and sQTL detected from 16 tissues jointly explain ~70% (SE=0.5%) of heritability, 44% more than expected from the same number of random variants, where trans e/sQTL contribute 24% (14% more than expected). Multi-tissue cis and trans e/sQTL also explain 71% (SE=0.3%) of heritability for the metabolome, demonstrating the essential role of proximal and distal regulatory variants in shaping mammalian phenotypes.

/pdf/gene-expression-and-rna-splicing-explain-large-proportions-2mma8ije.pdf

Gene expression and RNA splicing explain large proportions of the heritability for complex traits in cattle

To complete the genome-to-phenome map, transcriptome-wide association studies (TWAS) are performed to correlate genetically predicted gene expression with observed phenotypic measurements. However, the relatively small training population assayed with gene expression could limit the accuracy of TWAS. We propose Genetic Score Omics Regression (GSOR) correlating observed gene expression with genetically predicted phenotype, i.e., genetic score. The score, calculated using variants near genes with assayed expression, provides a powerful association test between cis-effects on gene expression and the trait. In simulated and real data, GSOR outperforms TWAS in detecting causal/informative genes. Applying GSOR to transcriptomes of 16 tissue (N∼5000) and 37 traits in ∼120,000 cattle, multi-trait meta-analyses of omics-associations (MTAO) found that, on average, each significant gene expression and splicing mediates cis-genetic effects on 8∼10 traits. Supported by Mendelian Randomisation, MTAO prioritised genes/splicing show increased evolutionary constraints. Many newly discovered genes/splicing regions underlie previously thought single-gene loci to influence multiple traits.

/pdf/genetic-score-omics-regression-and-multi-trait-meta-analysis-3ueuazns.pdf

Genetic score omics regression and multi-trait meta-analysis detect widespread cis-regulatory effects shaping bovine complex traits

Variants that change the expression of genes, i.e. expression quantitative trait loci or eQTL, provide an alternative to identify causal mutations. However, previous studies found little overlap between eQTL and conventional QTL. Here, we mapped and compared eQTL from blood and milk cells in ~400 dairy cattle with complex trait genome-wide association studies (GWAS) and Bayesian mixture models (BayesR) in ~103k cows. Increasing the sample size of eQTL mapping significantly increases the overlap between eQTL and QTL. As an example, a cis eQTL at Chr13:36387662(rs20846618) or the gene mohawk homeobox (MKX, peQTL=4.7e-08) significantly affects rump angle (‘pin set’: pGWAS=4.5e-10): a trait related to fertility and longevity. This eQTL is the top variant in BayesR mapping for rump angle. MKX is a transcription factor gene and is associated with tendon development and tail morphology in mice. Therefore, large-scale multi-omics fine-mapping can prioritise biologically sensible causal mutations for cattle complex traits. 

537. Large-scale ‘omics fine-mapping identifies expression quantitative trait loci significantly affecting cattle phenotypes

/pdf/correction-to-expression-of-mitochondrial-protein-genes-1h5labm3.pdf

B.A. Mason

Papers

Improving accuracy of genomic predictions within and between dairy cattle breeds with imputed high-density single nucleotide polymorphism panels

Gene expression and RNA splicing explain large proportions of the heritability for complex traits in cattle

Genetic score omics regression and multi-trait meta-analysis detect widespread cis-regulatory effects shaping bovine complex traits

537. Large-scale ‘omics fine-mapping identifies expression quantitative trait loci significantly affecting cattle phenotypes

Correction to: Expression of mitochondrial protein genes encoded by nuclear and mitochondrial genomes correlate with energy metabolism in dairy cattle