Phenome-wide genetic correlation analysis and genetically-informed causal inference of amyotrophic lateral sclerosis

Leveraging genome-wide association statistics generated from a large study of amyotrophic lateral sclerosis (ALS; 29,612 cases and 122,656 controls) and UK Biobank (UKB; 4,024 phenotypes, up to 361,194 participants), we conducted a phenome-wide genetic correlation analysis of ALS and identified 46 genetically-correlated traits, such as fluid intelligence score (rg = -0.21, p = 1.74x10-6), "spending time in pub or social club" (rg = 0.24, p = 2.77x10-6), non-work related walking (rg = -0.25, p = 1.95x10-6), college education (rg = -0.15, p = 7.08x10-5), "ever diagnosed with panic attacks (rg = 0.39, p = 4.24x10-5), and "self-reported other gastritis including duodenitis" (rg = 0.28, p = 1.4x10-3). To assess the putative directionality of these genetic correlations, we conducted a latent causal variable analysis, identifying significant genetic causality proportions (g[c]p) linking ALS to seven traits. While the genetic component of "self-reported other gastritis including duodenitis" showed a positive causal effect on ALS (g[c]p = 0.50, p = 1.26x10-29), the genetic liability to ALS is potentially causal for multiple traits, also including a positive effect on "ever being diagnosed with panic attacks" (g[c]p = 0.79, p = 5.011x10-15) and inverse effects on "other leisure/social group activities" (g[c]p = 0.66, p = 1x10-4) and prospective memory result (g[c]p = 0.35, p = 0.005). Our subsequent Mendelian randomization analysis indicated that some of these associations may be due to bi-directional effects. In conclusion, this is the first phenome-wide investigation of ALS polygenic architecture, highlighting the complex pleiotropy linking this disorder with different health domains.