Local genetic correlation quantifies the genetic similarity of complex traits in specific genomic regions, which could shed unique light on etiologic sharing and provide additional mechanistic insights into the genetic basis of complex traits compared to global genetic correlation. However, accurate estimation of local genetic correlation remains challenging, in part due to extensive linkage disequilibrium in local genomic regions and pervasive sample overlap across studies. We introduce SUPERGNOVA, a unified framework to estimate both global and local genetic correlations using summary statistics from genome-wide association studies. Through extensive simulations and analyses of 30 complex traits, we demonstrate that SUPERGNOVA substantially outperforms existing methods and identifies 150 trait pairs with significant local genetic correlations. In particular, we show that the positive, consistently-identified, 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. We believe that statistically-rigorous local genetic correlation analysis could accelerate progress in complex trait genetics research.

Genetic correlation is the correlation of phenotypic effects by genetic variants across the genome on two phenotypes. It is an informative metric to quantify the overall genetic similarity between complex traits, which provides insights into their polygenic genetic architecture. Several methods have been proposed to estimate genetic correlation based on data collected from genome-wide association studies (GWAS). Due to the easy access of GWAS summary statistics and computational efficiency, methods only requiring GWAS summary statistics as input have become more popular than methods utilizing individual-level genotype data. Here, we present a benchmark study for different summary-statistics-based genetic correlation estimation methods through simulation and real data applications. We focus on two major technical challenges in estimating genetic correlation: marker dependency caused by linkage disequilibrium (LD) and sample overlap between different studies. To assess the performance of different methods in the presence of these two challenges, we first conducted comprehensive simulations with diverse LD patterns and sample overlaps. Then we applied these methods to real GWAS summary statistics for a wide spectrum of complex traits. Based on these experiments, we conclude that methods relying on accurate LD estimation are less robust in real data applications due to the imprecision of LD obtained from reference panels. Our findings offer guidance on how to choose appropriate methods for genetic correlation estimation in post-GWAS analysis.

Genetic correlation is the correlation of additive genetic effects on two phenotypes. It is an informative metric to quantify the overall genetic similarity between complex traits, which provides insights into their polygenic genetic architecture. Several methods have been proposed to estimate genetic correlations based on data collected from genome- wide association studies (GWAS). Due to the easy access of GWAS summary statistics and computational efficiency, methods only requiring GWAS summary statistics as input have become more popular than methods utilizing individual-level genotype data. Here, we present a benchmark study for different summary-statistics-based genetic correlation estimation methods through simulation and real data applications. We focus on two major technical challenges in estimating genetic correlation: marker dependency caused by linkage disequilibrium (LD) and sample overlap between different studies. To assess the performance of different methods in the presence of these two challenges, we first conducted comprehensive simulations with diverse LD patterns and sample overlaps. Then we applied these methods to real GWAS summary statistics for a wide spectrum of complex traits. Based on these experiments, we conclude that methods relying on accurate LD estimation are less robust in real data applications compared to other methods due to the imprecision of LD obtained from reference panels. Our findings offer a guidance on how to appropriately choose the method for genetic correlation estimation in post-GWAS analysis in interpretation.

With the increasing accessibility of individual-level data from genome wide association studies, it is now common for researchers to have individual-level data of some traits in one specific population. For some traits, we can only access public released summary-level data due to privacy and safety concerns. The current methods to estimate genetic correlation can only be applied when the input data type of the two traits of interest is either both individual-level or both summary-level. When researchers have access to individual-level data for one trait and summary-level data for the other, they have to transform the individual-level data to summary-level data first and then apply summary data-based methods to estimate the genetic correlation. This procedure is computationally and statistically inefficient and introduces information loss. We introduce GENJI (Genetic correlation EstimatioN Jointly using Individual-level and summary data), a method that can estimate within-population or transethnic genetic correlation based on individual-level data for one trait and summary-level data for another trait. Through extensive simulations and analyses of real data on within-population and transethnic genetic correlation estimation, we show that GENJI produces more reliable and efficient estimation than summary data-based methods. Besides, when individual-level data are available for both traits, GENJI can achieve comparable performance than individual-level data-based methods. Downstream applications of genetic correlation can benefit from more accurate estimates. In particular, we show that more accurate genetic correlation estimation facilitates the predictability of cross-population polygenic risk scores.

With the increasing accessibility of individual-level data from genome wide association studies, it is now common for researchers to have individual-level data of some traits in one specific population. For some traits, we can only access public released summary-level data due to privacy and safety concerns. The current methods to estimate genetic correlation can only be applied when the input data type of the two traits of interest is either both individual-level or both summary-level. When researchers have access to individual-level data for one trait and summary-level data for the other, they have to transform the individual-level data to summary-level data first and then apply summary data-based methods to estimate the genetic correlation. This procedure is computationally and statistically inefficient and introduces information loss. We introduce GENJI (Genetic correlation EstimatioN Jointly using Individual-level and summary data), a method that can estimate within-population or transethnic genetic correlation based on individual-level data for one trait and summary-level data for another trait. Through extensive simulations and analyses of real data on within-population and transethnic genetic correlation estimation, we show that GENJI produces more reliable and efficient estimation than summary data-based methods. Besides, when individual-level data are available for both traits, GENJI can achieve comparable performance than individual-level data-based methods. Downstream applications of genetic correlation can benefit from more accurate estimates. In particular, we show that more accurate genetic correlation estimation facilitates the predictability of cross-population polygenic risk scores.