|
Abstract:
|
To date, more than one thousand genome-wide association studies (GWAS) have been completely. Due to the polygenic architecture of complex diseases, identification of risk single-nucleotide polymorphism (SNP) markers remains challenging. Until recently, most of conventional statistical methods only investigate one GWAS data set for one trait/disease at a time. To model the genetic correlation among complex diseases (formally known as "pleiotropy"), GPA [Chung et al., 2014] and EPS [Liu et al., 2016] used "four-group model" for two GWAS. The model complexity of parameter space grows exponentially as the number of GWAS increases. Here, we proposed LLR, the Latent Low Rank model to jointly analyze multiple GWAS. LLR uses a latent variable Z to indicate the null and non-null states as a ``two-group model" for each trait meanwhile the prior probability of the latent variable Z is modulated by a low-rank matrix X which is constrained by the nuclear norm. We applied LLR to jointly analyze p-values for 19 traits.
|
