LDpred2 is a widely used Bayesian method for building polygenic scores (PGSs). LDpred2-auto can infer the two parameters from the LDpred model, the SNP heritability h2 and polygenicity p, so that it does not require an additional validation dataset to choose best-performing parameters. The main aim of this paper is to properly validate the use of LDpred2-auto for inferring multiple genetic parameters. Here, we present a new version of LDpred2-auto that adds an optional third parameter α to its model, for modeling negative selection. We then validate the inference of these three parameters (or two, when using the previous model). We also show that LDpred2-auto provides per-variant probabilities of being causal that are well calibrated and can therefore be used for fine-mapping purposes. We also introduce a formula to infer the out-of-sample predictive performance r2 of the resulting PGS directly from the Gibbs sampler of LDpred2-auto. Finally, we extend the set of HapMap3 variants recommended to use with LDpred2 with 37% more variants to improve the coverage of this set, and we show that this new set of variants captures 12% more heritability and provides 6% more predictive performance, on average, in UK Biobank analyses.

Testing for associations in big data faces the problem of multiple comparisons, wherein true signals are difficult to detect on the background of all associations queried. This difficulty is particularly salient in human genetic association studies, in which phenotypic variation is often driven by numerous variants of small effect. The current strategy to improve power to identify these weak associations consists of applying standard marginal statistical approaches and increasing study sample sizes. Although successful, this approach does not leverage the environmental and genetic factors shared among the multiple phenotypes collected in contemporary cohorts. Here we developed covariates for multiphenotype studies (CMS), an approach that improves power when correlated phenotypes are measured on the same samples. Our analyses of real and simulated data provide direct evidence that correlated phenotypes can be used to achieve increases in power to levels often surpassing the power gained by a twofold increase in sample size.

Although the cohort-level accuracy of polygenic risk scores (PRSs)-estimates of genetic value at the individual level-has been widely assessed, uncertainty in PRSs remains underexplored. In the present study, we show that Bayesian PRS methods can estimate the variance of an individual's PRS and can yield well-calibrated credible intervals via posterior sampling. For 13 real traits in the UK Biobank (n = 291,273 unrelated 'white British'), we observe large variances in individual PRS estimates which impact interpretation of PRS-based stratification; averaging across traits, only 0.8% (s.d. = 1.6%) of individuals with PRS point estimates in the top decile have corresponding 95% credible intervals fully contained in the top decile. We provide an analytical estimator for the expectation of individual PRS variance as a function of SNP heritability, number of causal SNPs and sample size. Our results showcase the importance of incorporating uncertainty in individual PRS estimates into subsequent analyses.

Recent work has shown that much of the missing heritability of complex traits can be resolved by estimates of heritability explained by all genotyped SNPs. However, it is currently unknown how much heritability is missing due to poor tagging or additional causal variants at known GWAS loci. Here, we use variance components to quantify the heritability explained by all SNPs at known GWAS loci in nine diseases from WTCCC1 and WTCCC2. After accounting for expectation, we observed all SNPs at known GWAS loci to explain 1.29 x more heritability than GWAS-associated SNPs on average (P=3.3 x 10⁻⁵). For some diseases, this increase was individually significant: 2.07 x for Multiple Sclerosis (MS) (P=6.5 x 10⁻⁹) and 1.48 x for Crohns Disease (CD) (P = 1.3 x 10⁻³); all analyses of autoimmune diseases excluded the well-studied MHC region. Additionally, we found that GWAS loci from other related traits also explained significant heritability. The union of all autoimmune disease loci explained 7.15 x more MS heritability than known MS SNPs (P < 1.0 x 10⁻¹⁶ and 2.20 x more CD heritability than known CD SNPs (P = 6.1 x 10⁻⁹), with an analogous increase for all autoimmune diseases analyzed. We also observed significant increases in an analysis of > 20,000 Rheumatoid Arthritis (RA) samples typed on ImmunoChip, with 2.37 x more heritability from all SNPs at GWAS loci (P = 2.3 x 10⁻⁶) and 5.33 x more heritability from all autoimmune disease loci (P < 1 x 10⁻¹⁶ compared to known RA SNPs (including those identified in this cohort). Our methods adjust for LD between SNPs, which can bias standard estimates of heritability from SNPs even if all causal variants are typed. By comparing adjusted estimates, we hypothesize that the genome-wide distribution of causal variants is enriched for low-frequency alleles, but that causal variants at known GWAS loci are skewed towards common alleles. These findings have important ramifications for fine-mapping study design and our understanding of complex disease architecture.

Polygenic scores (PGSs) have limited portability across different groupings of individuals (for example, by genetic ancestries and/or social determinants of health), preventing their equitable use^1-3. PGS portability has typically been assessed using a single aggregate population-level statistic (for example, R²)⁴, ignoring inter-individual variation within the population. Here, using a large and diverse Los Angeles biobank⁵ (ATLAS, n = 36,778) along with the UK Biobank⁶ (UKBB, n = 487,409), we show that PGS accuracy decreases individual-to-individual along the continuum of genetic ancestries⁷ in all considered populations, even within traditionally labelled 'homogeneous' genetic ancestries. The decreasing trend is well captured by a continuous measure of genetic distance (GD) from the PGS training data: Pearson correlation of -0.95 between GD and PGS accuracy averaged across 84 traits. When applying PGS models trained on individuals labelled as white British in the UKBB to individuals with European ancestries in ATLAS, individuals in the furthest GD decile have 14% lower accuracy relative to the closest decile; notably, the closest GD decile of individuals with Hispanic Latino American ancestries show similar PGS performance to the furthest GD decile of individuals with European ancestries. GD is significantly correlated with PGS estimates themselves for 82 of 84 traits, further emphasizing the importance of incorporating the continuum of genetic ancestries in PGS interpretation. Our results highlight the need to move away from discrete genetic ancestry clusters towards the continuum of genetic ancestries when considering PGSs.

Linear mixed models are a powerful statistical tool for identifying genetic associations and avoiding confounding. However, existing methods are computationally intractable in large cohorts and may not optimize power. All existing methods require time cost O(MN(2)) (where N is the number of samples and M is the number of SNPs) and implicitly assume an infinitesimal genetic architecture in which effect sizes are normally distributed, which can limit power. Here we present a far more efficient mixed-model association method, BOLT-LMM, which requires only a small number of O(MN) time iterations and increases power by modeling more realistic, non-infinitesimal genetic architectures via a Bayesian mixture prior on marker effect sizes. We applied BOLT-LMM to 9 quantitative traits in 23,294 samples from the Women's Genome Health Study (WGHS) and observed significant increases in power, consistent with simulations. Theory and simulations show that the boost in power increases with cohort size, making BOLT-LMM appealing for genome-wide association studies in large cohorts.

Despite recent progress on estimating the heritability explained by genotyped SNPs (h(2)g), a large gap between h(2)g and estimates of total narrow-sense heritability (h(2)) remains. Explanations for this gap include rare variants or upward bias in family-based estimates of h(2) due to shared environment or epistasis. We estimate h(2) from unrelated individuals in admixed populations by first estimating the heritability explained by local ancestry (h(2)γ). We show that h(2)γ = 2FSTCθ(1 - θ)h(2), where FSTC measures frequency differences between populations at causal loci and θ is the genome-wide ancestry proportion. Our approach is not susceptible to biases caused by epistasis or shared environment. We applied this approach to the analysis of 13 phenotypes in 21,497 African-American individuals from 3 cohorts. For height and body mass index (BMI), we obtained h(2) estimates of 0.55 ± 0.09 and 0.23 ± 0.06, respectively, which are larger than estimates of h(2)g in these and other data but smaller than family-based estimates of h(2).

Depression is a common psychiatric disorder and a leading cause of disability worldwide. Here we conducted a genome-wide association study meta-analysis of six datasets, including >1.3 million individuals (371,184 with depression) and identified 243 risk loci. Overall, 64 loci were new, including genes encoding glutamate and GABA receptors, which are targets for antidepressant drugs. Intersection with functional genomics data prioritized likely causal genes and revealed new enrichment of prenatal GABAergic neurons, astrocytes and oligodendrocyte lineages. We found depression to be highly polygenic, with ~11,700 variants explaining 90% of the single-nucleotide polymorphism heritability, estimating that >95% of risk variants for other psychiatric disorders (anxiety, schizophrenia, bipolar disorder and attention deficit hyperactivity disorder) were influencing depression risk when both concordant and discordant variants were considered, and nearly all depression risk variants influenced educational attainment. Additionally, depression genetic risk was associated with impaired complex cognition domains. We dissected the genetic and clinical heterogeneity, revealing distinct polygenic architectures across subgroups of depression and demonstrating significantly increased absolute risks for recurrence and psychiatric comorbidity among cases of depression with the highest polygenic burden, with considerable sex differences. The risks were up to 5- and 32-fold higher than cases with the lowest polygenic burden and the background population, respectively. These results deepen the understanding of the biology underlying depression, its disease progression and inform precision medicine approaches to treatment.

Regulatory and coding variants are known to be enriched with associations identified by genome-wide association studies (GWASs) of complex disease, but their contributions to trait heritability are currently unknown. We applied variance-component methods to imputed genotype data for 11 common diseases to partition the heritability explained by genotyped SNPs (hg(2)) across functional categories (while accounting for shared variance due to linkage disequilibrium). Extensive simulations showed that in contrast to current estimates from GWAS summary statistics, the variance-component approach partitions heritability accurately under a wide range of complex-disease architectures. Across the 11 diseases DNaseI hypersensitivity sites (DHSs) from 217 cell types spanned 16% of imputed SNPs (and 24% of genotyped SNPs) but explained an average of 79% (SE = 8%) of hg(2) from imputed SNPs (5.1× enrichment; p = 3.7 × 10(-17)) and 38% (SE = 4%) of hg(2) from genotyped SNPs (1.6× enrichment, p = 1.0 × 10(-4)). Further enrichment was observed at enhancer DHSs and cell-type-specific DHSs. In contrast, coding variants, which span 1% of the genome, explained <10% of hg(2) despite having the highest enrichment. We replicated these findings but found no significant contribution from rare coding variants in independent schizophrenia cohorts genotyped on GWAS and exome chips. Our results highlight the value of analyzing components of heritability to unravel the functional architecture of common disease.

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.