Patrick Turley

We introduce a novel approach, Genomic-Relatedness-Matched Association (GRMA) studies, as an alternative to genome-wide association studies (GWAS). GWAS are typically restricted to samples of mostly unrelated individuals with a single, shared continental ancestry and nevertheless can still be biased by gene-environment correlation and assortative mating. In contrast, GRMA can be implemented in ancestrally diverse samples—retaining individuals of mixed or underrepresented ancestries and eliminating the need to assign labels to ancestry groups—and can reduce bias relative to standard GWAS. GRMA matches each individual to a group of controls whose pairwise relatedness with the individual exceeds a user-specified threshold. It generates SNP-level summary statistics based on within-group associations. In applications using the UK Biobank and All of Us data, we find that GRMA compares favorably to GWAS methods in terms of bias, precision, and population coverage. GRMA enables several novel findings; for example, we find that “genetic nurture” is unlikely to be an important source of genome-wide bias in population GWAS of body mass index, height, and educational attainment. The method is computationally efficient and supported by open-source software, facilitating its application in large-scale scientific and health-related studies.

Polygenic indexes (PGIs) — DNA-based predictors of individual phenotypes — have become essential tools across biomedical and social sciences. We introduce Version 2 of the Polygenic Index Repository, which expands phenotype coverage from 47 to 61, increases the number of participating datasets from 11 to 20, and adopts a more consistent and improved methodology for PGI construction. For 16 phenotypes, we leverage summary statistics from an updated GWAS meta-analysis with greater statistical power compared to the original release, thereby improving the PGI’s predictive power. To improve power for family-based analyses, we provide imputed parental PGIs in all datasets with first-degree relatives and offer a framework for interpreting results from analyses that control for parental PGIs. We illustrate the utility of parental PGIs using two applications: (1) comparing PGI associations with and without parental PGI controls for all phenotypes in two Repository datasets with family data, and (2) for BMI and diastolic blood pressure, exploring the contribution of causal versus non-causal components of PGI associations to the imperfect portability of PGIs across subgroups within a genetic ancestry. Collectively, the updates enhance predictive performance, broaden the Repository’s scope, and introduce novel resources that reduce confounding bias and improve interpretability.

Polygenic indexes (PGIs) trained on samples of European genetic ancestries often lose substantial predictive power when applied to non-European ancestries. While this portability problem is well recognized, its manifestation in behavioral and social traits remains understudied, and the factors driving this accuracy loss warrant more comprehensive analysis. Using data from the UK Biobank and Health and Retirement Study, we conduct a systematic analysis of PGI portability for 52 health-related, behavioral, and social phenotypes. We advance prior literature by using genome-wide PGIs, assessing cross-ancestry heritability differences, and comparing the performance of PGIs based on standard versus family-based GWAS. Our findings confirm systematic reductions in PGI predictive power for non-European ancestries—lowest in African (24%), followed by East Asian (37%) and South Asian (51%) genetic ancestries—with biologically proximal traits exhibiting greater portability than behavioral and social traits. We show that the relative importance of factors underlying reduced portability varies across traits and ancestries: in African ancestries, linkage disequilibrium and allele frequency differences explain most of the loss (82%), compared with smaller contributions in East (34%) and South Asian (25%) ancestries. Finally, we find that family-based GWAS PGIs can modestly improve portability for select traits, such as BMI in African ancestry, suggesting that part of the portability gap may reflect population-specific confounds in standard PGIs.

Family-based genome-wide association studies (FGWASs) use random, within-family genetic variation to remove confounding from estimates of direct genetic effects (DGEs). Here we introduce a ‘unified estimator’ that includes individuals without genotyped relatives, unifying standard and FGWAS while increasing power for DGE estimation. We also introduce a ‘robust estimator’ that is not biased in structured and/or admixed populations. In an analysis of 19 phenotypes in the UK Biobank, the unified estimator in the White British subsample and the robust estimator (applied without ancestry restrictions) increased the effective sample size for DGEs by 46.9% to 106.5% and 10.3% to 21.0%, respectively, compared to using genetic differences between siblings. Polygenic predictors derived from the unified estimator demonstrated superior out-of-sample prediction ability compared to other family-based methods. We implemented the methods in the software package snipar in an efficient linear mixed model that accounts for sample relatedness and sibling shared environment.

We develop a method, SBayesRC, that integrates genome-wide association study (GWAS) summary statistics with functional genomic annotations to improve polygenic prediction of complex traits. Our method is scalable to whole-genome variant analysis and refines signals from functional annotations by allowing them to affect both causal variant probability and causal effect distribution. We analyze 50 complex traits and diseases using ∼7 million common single-nucleotide polymorphisms (SNPs) and 96 annotations. SBayesRC improves prediction accuracy by 14% in European ancestry and up to 34% in cross-ancestry prediction compared to the baseline method SBayesR, which does not use annotations, and outperforms other methods, including LDpred2, LDpred-funct, MegaPRS, PolyPred-S and PRS-CSx. Investigation of factors affecting prediction accuracy identifies a significant interaction between SNP density and annotation information, suggesting whole-genome sequence variants with annotations may further improve prediction. Functional partitioning analysis highlights a major contribution of evolutionary constrained regions to prediction accuracy and the largest per-SNP contribution from nonsynonymous SNPs.

Gestational diabetes mellitus (GDM) is a common metabolic disorder affecting more than 16 million pregnancies annually worldwide^1,2. GDM is related to an increased lifetime risk of type 2 diabetes (T2D)^1–3, with over a third of women developing T2D within 15 years of their GDM diagnosis. The diseases are hypothesized to share a genetic predisposition^1–7, but few studies have sought to uncover the genetic underpinnings of GDM. Most studies have evaluated the impact of T2D loci only^8–10, and the three prior genome-wide association studies of GDM^11–13 have identified only five loci, limiting the power to assess to what extent variants or biological pathways are specific to GDM. We conducted the largest genome-wide association study of GDM to date in 12,332 cases and 131,109 parous female controls in the FinnGen study and identified 13 GDM-associated loci, including nine new loci. Genetic features distinct from T2D were identified both at the locus and genomic scale. Our results suggest that the genetics of GDM risk falls into the following two distinct categories: one part conventional T2D polygenic risk and one part predominantly influencing mechanisms disrupted in pregnancy. Loci with GDM-predominant effects map to genes related to islet cells, central glucose homeostasis, steroidogenesis and placental expression.

Background: Recent genetics studies for Alzheimer’s disease have identified multiple risk variants affecting amyloid processing, immune response, lipid transport, and endocytosis, among others 1. More than half the risk genes associated with late‐onset AD are selectively expressed in microglia and peripheral myeloid cells. Yet, we know little about the underlying biology or how myeloid cells contribute to AD pathogenesis.Method: To gain insights into the broad impact of genetics on microglia, we selected 50 cell lines from the CIRM biobank with variable APOE genotypes and polygenic risk scores (PRS). Building on our in vitro platform that recapitulates microglia states in vitro 2, we used villages 3, or pooled culture, to differentiate these lines into microglia and assessed key microglia functions, transcriptional profiles, and response to various perturbations.Result: Cell villages allow us to compare phenotypes across lines with little technical variability robustly. Our data determined functional phenotypes that were altered by PRS and others that were affected by APOE genotypes. Moreover, quantitative trait locus (QTL) analysis identified new variants affecting phagocytosis, proliferation, and response to lipids. Finally, single‐cell RNA sequencing of our village pinpointed key microglia pathways altered between high and low PRS individuals.Conclusion: Cell village allows for the characterization of up to hundreds of cell lines while minimizing technical variability allowing us to identify phenotypes dysregulated across various genetic backgrounds. Our data pinpointed the impact of a novel Alzheimer’s disease PRS on microglia states and functions and identified new variants affecting microglia functions.1. Sims, R., Hill, M. & Williams, J. Nat Neurosci 23, 311‐322 (2020). 2. Dolan, M.‐J. et al. Biorxiv 2022.05.02.490100 (2022) https://doi.org/10.1101/2022.05.02.490100. 3. Wells, M. F. et al. Biorxiv 2021.11.08.467815 (2021) https://doi.org/10.1101/2021.11.08.467815.

The effects of assortative mating (AM) on estimates from genetic studies has been receiving increasing attention in recent years. We extend existing AM theory to more general models of sorting and conclude that correct theory-based AM adjustments require knowledge of complicated, unknown historical sorting patterns. We propose a simple, general-purpose approach using polygenic indexes (PGIs). Our approach can estimate the fraction of genetic variance and genetic correlation that is driven by AM. Our approach is less effective when applied to Mendelian randomization (MR) studies for two reasons: AM can induce a form of selection bias in MR studies that remains after our adjustment; and, in the MR context, the adjustment is particularly sensitive to PGI estimation error. Using data from the UK Biobank, we find that AM inflates genetic correlation estimates between health traits and education by 14% on average. Our results suggest caution in interpreting genetic correlations or MR estimates for traits subject to AM.

Effects estimated by genome-wide association studies (GWASs) include effects of alleles in an individual on that individual (direct genetic effects), indirect genetic effects (for example, effects of alleles in parents on offspring through the environment) and bias from confounding. Within-family genetic variation is random, enabling unbiased estimation of direct genetic effects when parents are genotyped. However, parental genotypes are often missing. We introduce a method that imputes missing parental genotypes and estimates direct genetic effects. Our method, implemented in the software package snipar (single-nucleotide imputation of parents), gives more precise estimates of direct genetic effects than existing approaches. Using 39,614 individuals from the UK Biobank with at least one genotyped sibling/parent, we estimate the correlation between direct genetic effects and effects from standard GWASs for nine phenotypes, including educational attainment (r = 0.739, standard error (s.e.) = 0.086) and cognitive ability (r = 0.490, s.e. = 0.086). Our results demonstrate substantial confounding bias in standard GWASs for some phenotypes.

Polygenic indexes (PGIs) are DNA-based predictors. Their value for research in many scientific disciplines is growing rapidly. As a resource for researchers, we used a consistent methodology to construct PGIs for 47 phenotypes in 11 datasets. To maximize the PGIs’ prediction accuracies, we constructed them using genome-wide association studies—some not previously published—from multiple data sources, including 23andMe and UK Biobank. We present a theoretical framework to help interpret analyses involving PGIs. A key insight is that a PGI can be understood as an unbiased but noisy measure of a latent variable we call the ‘additive SNP factor’. Regressions in which the true regressor is this factor but the PGI is used as its proxy therefore suffer from errors-in-variables bias. We derive an estimator that corrects for the bias, illustrate the correction, and make a Python tool for implementing it publicly available.