SUPERGNOVA: local genetic correlation analysis reveals heterogeneous etiologic sharing of complex traits

Yiliang Zhang,Kunling Huang,Donna M Werling,Yuchang Wu,Qiongshi Lu,Boyang Li,Hongyu Zhao,James J Li,Brittany G Travers,Zhaolong Yu,Wei Liu,Yixuan Ye,Xiaoyuan Zhong

doi:10.1186/s13059-021-02478-w

Abstract

Local genetic correlation quantifies the genetic similarity of complex traits in specific genomic regions. However, accurate estimation of local genetic correlation remains challenging, due to linkage disequilibrium in local genomic regions and sample overlap across studies. We introduce SUPERGNOVA, a statistical framework to estimate local genetic correlations using summary statistics from genome-wide association studies. We demonstrate that SUPERGNOVA outperforms existing methods through simulations and analyses of 30 complex traits. In particular, we show that the positive yet paradoxical genetic correlation between autism spectrum disorder and cognitive performance could be explained by two etiologically distinct genetic signatures with bidirectional local genetic correlations.

Highlights

Genome-wide association study (GWAS) has achieved remarkable success in the past 15 years and has identified numerous single-nucleotide polymorphisms (SNPs) associated with complex human traits and diseases [1]
We demonstrate that SUPERGNOVA provides statistically rigorous and computationally efficient inference for both global and local genetic correlations and substantially outperforms existing methods when applied to local genomic regions
Ubiquitous sample overlap across GWASs introduces additional correlations among association statistics from different studies, which further complicates the estimation of genetic correlation

Summary

Introduction

Genome-wide association study (GWAS) has achieved remarkable success in the past 15 years and has identified numerous single-nucleotide polymorphisms (SNPs) associated with complex human traits and diseases [1]. Accessible summary statistics from GWAS, in conjunction with advances in analytical methods that use marginal association statistics as input, have circumvented logistical challenges in data sharing and greatly accelerated research in complex trait genetics [2]. With these advancements, multi-trait modeling has undergone rapid developments, leading to the emergence of numerous methods that study the shared genetic basis across multiple phenotypes [3,4,5,6,7,8].

Objectives

Methods

Results

Discussion

Conclusion