Variable prediction accuracy of polygenic scores within an ancestry group

Three legs of the missing heritability problem

The so-called 'missing heritability problem' is often characterized by behavior geneticists as a numerical discrepancy between alternative kinds of heritability. For example, while 'traditional heritability' derived from twin and family studies indicates that approximately ∼50% of variation in intelligence is attributable to genetics, 'SNP heritability' derived from genome-wide association studies indicates that only ∼10% of variation in intelligence is attributable to genetics. This 40% gap in variance accounted for by alternative kinds of heritability is frequently referred to as what's "missing." Philosophers have picked up on this reading, suggesting that "dissolving" the missing heritability problem is merely a matter of closing the numerical gap between traditional and molecular kinds of heritability. We argue that this framing of the problem undervalues the severity of the many challenges to scientific understanding of the "heritability" of human behavior. On our view, resolving the numerical discrepancies between alternative kinds of heritability will do little to advance scientific explanation and understanding of behavior genetics. Thus, we propose a new conceptual framework of the missing heritability problem that comprises three independent methodological and explanatory challenges: the numerical gap, the prediction gap, and the mechanism gap.

Half a century later and we're back where we started: How the problem of locality turned in to the problem of portability

In the 1970s, Lewontin sparked a debate about a problem of locality, by making the case that any given heritability estimate is local to the original population and environment studied, and could not be generalized to other populations and environments. Nearly 50 years later, a new problem of portability has emerged: the predictive accuracy of polygenic scores diminishes when applied to populations whose characteristics are different from the original population sample. This paper briefly reviews the nature of each problem and analyzes their similarities and differences in three areas: 1) conceptual underpinnings, 2) causal explanations, and 3) practical, social, and political implications. Although conceptually and methodologically different from the problem of locality in important respects, the problem of portability facing contemporary genomics today should come as no surprise, as it is an inevitable outcome of the kinds of problematic inferences detailed by Lewontin nearly half a century ago.

A model and test for coordinated polygenic epistasis in complex traits

Interactions between genetic variants-epistasis-is pervasive in model systems and can profoundly impact evolutionary adaption, population disease dynamics, genetic mapping, and precision medicine efforts. In this work, we develop a model for structured polygenic epistasis, called coordinated epistasis (CE), and prove that several recent theories of genetic architecture fall under the formal umbrella of CE. Unlike standard epistasis models that assume epistasis and main effects are independent, CE captures systematic correlations between epistasis and main effects that result from pathway-level epistasis, on balance skewing the penetrance of genetic effects. To test for the existence of CE, we propose the even-odd (EO) test and prove it is calibrated in a range of realistic biological models. Applying the EO test in the UK Biobank, we find evidence of CE in 18 of 26 traits spanning disease, anthropometric, and blood categories. Finally, we extend the EO test to tissue-specific enrichment and identify several plausible tissue-trait pairs. Overall, CE is a dimension of genetic architecture that can capture structured, systemic forms of epistasis in complex human traits.

Schizophrenia Polygenic Risk and Brain Structural Changes in Methamphetamine-Associated Psychosis in a South African Population

Background

The genetic architecture of psychotic disorders is complex, with hundreds of genetic risk loci contributing to a polygenic model of disease. Overlap in the genetics of psychotic disorders and brain measures has been found in European populations, but has not been explored in populations of African ancestry. The aim of this study was to determine whether a relationship exists between a schizophrenia-derived PRS and (i) methamphetamine associated psychosis (MAP), and (ii) brain structural measures, in a South African population.

Methods

The study sample consisted of three participant groups: 31 individuals with MAP, 48 with apsychotic methamphetamine dependence, and 49 healthy controls. Using PRSice, PRS was generated for each of the participants with GWAS summary statistics from the Psychiatric Genomics Consortium Schizophrenia working group (PGC-SCZ2) as the discovery dataset. Regression analyses were performed to determine associations of PRS, with diagnosis, whole brain, and regional gray and white matter measures.

Results

Schizophrenia-derived PRS did not significantly predict MAP diagnosis. After correction for multiple testing, no significant associations were found between PRS and brain measures across all groups.

Discussion

The lack of significant associations here may indicate that the study is underpowered, that brain volumes in MAP are due to factors other than polygenic risk for schizophrenia, or that PRS derived from a largely European discovery set has limited utility in individuals of African ancestry. Larger studies, that include diverse populations, and more nuanced brain measures, may help elucidate the relationship between schizophrenia-PRS, brain structural changes, and psychosis.

Conclusion

This research presents the first PRS study to investigate shared genetic effects across psychotic disorders and brain structural measures in an African population. Ancestrally comparable discovery datasets may be useful for future African genetic research.

Leveraging correlations between variants in polygenic risk scores to detect heterogeneity in GWAS cohorts

Evidence from both GWAS and clinical observation has suggested that certain psychiatric, metabolic, and autoimmune diseases are heterogeneous, comprising multiple subtypes with distinct genomic etiologies and Polygenic Risk Scores (PRS). However, the presence of subtypes within many phenotypes is frequently unknown. We present CLiP (Correlated Liability Predictors), a method to detect heterogeneity in single GWAS cohorts. CLiP calculates a weighted sum of correlations between SNPs contributing to a PRS on the case/control liability scale. We demonstrate mathematically and through simulation that among i.i.d. homogeneous cases generated by a liability threshold model, significant anti-correlations are expected between otherwise independent predictors due to ascertainment on the hidden liability score. In the presence of heterogeneity from distinct etiologies, confounding by covariates, or mislabeling, these correlation patterns are altered predictably. We further extend our method to two additional association study designs: CLiP-X for quantitative predictors in applications such as transcriptome-wide association, and CLiP-Y for quantitative phenotypes, where there is no clear distinction between cases and controls. Through simulations, we demonstrate that CLiP and its extensions reliably distinguish between homogeneous and heterogeneous cohorts when the PRS explains as low as 3% of variance on the liability scale and cohorts comprise 50, 000 − 100, 000 samples, an increasingly practical size for modern GWAS. We apply CLiP to heterogeneity detection in schizophrenia cohorts totaling > 50, 000 cases and controls collected by the Psychiatric Genomics Consortium. We observe significant heterogeneity in mega-analysis of the combined PGC data (p-value 8.54 × 0⁻⁴), as well as in individual cohorts meta-analyzed using Fisher’s method (p-value 0.03), based on significantly associated variants. We also apply CLiP-Y to detect heterogeneity in neuroticism in over 10, 000 individuals from the UK Biobank and detect heterogeneity with a p-value of 1.68 × 10⁻⁹. Scores were not significantly reduced when partitioning by known subclusters (“Depression” and “Worry”), suggesting that these factors are not the primary source of observed heterogeneity.

Several traits, such as bipolar disease, are known to be heterogeneous and comprise distinct subtypes with unique genomic associations. For other traits such as schizophrenia, heterogeneity may be suspected, but specific subtypes are less well characterized. Furthermore, conventional mixture model-based methods to detect subtypes in genome-wide association data struggle with the high polygenicity of complex traits. We propose CLiP (Correlated Liability Predictors), a method that does not identify subtype-specific effects, but is very well-powered to detect heterogeneity of any kind within the very weak signals of GWAS. CLiP serves as a method to flag particular phenotypes for potential further study into the genomic factors driving heterogeneity, as well as a means to evaluate the transferability of polygenic risk scores. We also develop extensions of CLiP applicable to scoring heterogeneity in quantitative phenotypes and quantitative predictors such as gene expression. We apply CLiP to scoring heterogeneity in schizophrenia cohorts from the Psychiatric Genomics Consortium and neuroticism in individuals in the UK Biobank and find significant heterogeneity in both phenotypes, warranting further study.

Polygenic scores for psychiatric disease: from research tool to clinical application

Propensity to psychiatric disease involves the contribution of multiple genetic variants with small individual effects (i. e., polygenicity). This contribution can be summarized using polygenic scores (PGSs). The present article discusses the methodological foundations of PGS calculation, together with the limitations and caveats of their use. Furthermore, we show that in terms of using genetic information to address the complexities of mental disorders, PGSs have become a standard tool in psychiatric research. PGS also have the potential for translation into clinical practice. Although PGSs alone do not allow reliable disease prediction, they have major potential value in terms of risk stratification, the identification of disorder subtypes, functional investigations, and case selection for experimental models. However, given the stigma associated with mental illness and the limited availability of effective interventions, risk prediction for common psychiatric disorders must be approached with particular caution, particularly in the non-regulated consumer context.

Evaluating the promise of inclusion of African ancestry populations in genomics

The lack of representation of diverse ancestral backgrounds in genomic research is well-known, and the resultant scientific and ethical limitations are becoming increasingly appreciated. The paucity of data on individuals with African ancestry is especially noteworthy as Africa is the birthplace of modern humans and harbors the greatest genetic diversity. It is expected that greater representation of those with African ancestry in genomic research will bring novel insights into human biology, and lead to improvements in clinical care and improved understanding of health disparities. Now that major efforts have been undertaken to address this failing, is there evidence of these anticipated advances? Here, we evaluate the promise of including diverse individuals in genomic research in the context of recent literature on individuals of African ancestry. In addition, we discuss progress and achievements on related technological challenges and diversity among scientists conducting genomic research.

Influence of Genetic Interactions on Polygenic Prediction

Prediction of phenotypes from genotypes is an important objective to fulfill the promises of genomics, precision medicine and agriculture. Although it's now possible to account for the majority of genetic variation through model fitting, prediction of phenotypes remains a challenge, especially across populations that have diverged in the past. In this study, we designed simulation experiments to specifically investigate the role of genetic interactions in failure of polygenic prediction. We found that non-additive genetic interactions can significantly reduce the accuracy of polygenic prediction. Our study demonstrated the importance of considering genetic interactions in genetic prediction.

Evolutionary perspectives on polygenic selection, missing heritability, and GWAS

Genome-wide association studies (GWAS) have successfully identified many trait-associated variants, but there is still much we do not know about the genetic basis of complex traits. Here, we review recent theoretical and empirical literature regarding selection on complex traits to argue that "missing heritability" is as much an evolutionary problem as it is a statistical problem. We discuss empirical findings that suggest a role for selection in shaping the effect sizes and allele frequencies of causal variation underlying complex traits, and the limitations of these studies. We then use simulations of selection, realistic genome structure, and complex human demography to illustrate the results of recent theoretical work on polygenic selection, and show that statistical inference of causal loci is sharply affected by evolutionary processes. In particular, when selection acts on causal alleles, it hampers the ability to detect causal loci and constrains the transferability of GWAS results across populations. Last, we discuss the implications of these findings for future association studies, and suggest that future statistical methods to infer causal loci for genetic traits will benefit from explicit modeling of the joint distribution of effect sizes and allele frequencies under plausible evolutionary models.

Deconstructing the sources of genotype-phenotype associations in humans

Efforts to link variation in the human genome to phenotypes have progressed at a tremendous pace in recent decades. Most human traits have been shown to be affected by a large number of genetic variants across the genome. To interpret these associations and to use them reliably-in particular for phenotypic prediction-a better understanding of the many sources of genotype-phenotype associations is necessary. We summarize the progress that has been made in this direction in humans, notably in decomposing direct and indirect genetic effects as well as population structure confounding. We discuss the natural next steps in data collection and methodology development, with a focus on what can be gained by analyzing genotype and phenotype data from close relatives.