Leveraging correlations between polygenic risk score predictors to detect heterogeneity in GWAS cohorts

Abstract

9Evidence from both GWAS and clinical observation has suggested that certain psychiatric, metabolic, and 10 autoimmune diseases are heterogeneous, comprising multiple subtypes with distinct genomic etiologies and 11 Polygenic Risk Scores (PRS). However, the presence of subtypes within many phenotypes is frequently 12 unknown. We present CLiP (Correlated Liability Predictors), a method to detect heterogeneity in single 13 GWAS cohorts. CLiP calculates a weighted sum of correlations between SNPs contributing to a PRS on 14 the case/control liability scale. We demonstrate mathematically and through simulation that among i.i.d. 15 homogeneous cases, significant anti-correlations are expected between otherwise independent predictors due 16 to ascertainment on the hidden liability score. In the presence of heterogeneity from distinct etiologies, 17 confounding by covariates, or mislabeling, these correlation patterns are altered predictably. We further 18 extend our method to two additional association study designs: CLiP-X for quantitative predictors in 19 applications such as transcriptome-wide association, and CLiP-Y for quantitative phenotypes, where there 20 is no clear distinction between cases and controls. Through simulations, we demonstrate that CLiP and its 21 extensions reliably distinguish between homogeneous and heterogeneous cohorts when the PRS explains as 22 low as 5% of variance on the liability scale and cohorts comprise 50, 000 − 100, 000 samples, an increasingly 23 practical size for modern GWAS. We apply CLiP to heterogeneity detection in schizophrenia cohorts totaling 24 > 50, 000 cases and controls collected by the Psychiatric Genomics Consortium. We observe significant 25 heterogeneity in mega-analysis of the combined PGC data (p-value 8.54e-4), as well as in individual cohorts 26 meta-analyzed using Fisher's method (p-value 0.03), based on significantly associated variants. 27 2 1 Introduction 28In recent years Genome-Wide Association Studies (GWAS) have identified thousands of genomic risk factors 29 and generated insights into disease etiologies and potential treatments [1, 2, 3]. Increasingly, there has been 30 interest in advancing beyond these associations towards obtaining a deeper understanding the mechanisms 31 by which genomic factors influence disease [1, 4]. These require models beyond simply combining linear 32 effects of variants, as they often modulate phenotypes indirectly, though the expression of other genes [5, 6]. 33 One such avenue has concerned the apparent heterogeneity of diseases which has not been sufficiently 34 recognized by GWAS: while individuals in cohorts for these studies are frequently classified simply as cases or 35 controls, clinical evidence for several GWAS traits have suggested that there are multiple different subtypes 36 of diseases consisting of distinct sets of symptoms and association with distinct rare risk alleles [7, 8]. 37For example, polygenic risk scores for major depressive disorder explain more of the phenotypic variance 38 when ca...

Interaction Testing and Polygenic Risk Scoring to Estimate the Association of Common Genetic Variants With Treatment Resistance in Schizophrenia

“…27,28 Some of the largest cohorts in these research initiatives were recruited based on clozapine prescription (a proxy of TRS status), and forming a case-case data set from them would require avoiding confounding factors, such as GWAS batch effects 30 or population stratification, 31 which are difficult to control in a multiple-cohort design. 32 As a safeguard against these, we have used a meta-analytic procedure to assess the differences between GWAS in which individuals with TRS and non-TRS have been compared with matched sets of healthy controls, before comparing the allelic association effect sizes of these 2 GWASs on a genomewide basis to create a GWAS specific to treatment resistance.…”