There are several reasons these GEITP pages continue to share publications dealing with genome-wide association studies (GWAS). One, GWAS are becoming very common (due to lower costs, efficiency of carrying out). Two, GWAS are including much larger sample sizes (of DNA, and of information collected on medical history, family history, and answers to detailed questionnaires) — in an increasing number of databases, worldwide; these will find increasingly larger numbers of very-small-effect genes contributing to a trait. Three, additional statistical analyses are now being added [e.g. linkage disequilibrium score regression (LDSC), N-weighted multivariate analysis of GWAS (N-GWAMA), and model-averaging GWAMA (MA-GWAMA) — as detailed in the attached report] to the original (simple) straightforward GWAS. And finally, these methods have begun to be used in pharmacology and environmental toxicology — except, obviously, it is much more difficult to find even hundreds of patients on a particular drug, class of drugs, or having a specific adverse drug reaction; it is also much more difficult to find hundreds of workers exposed to a particular environmental toxicant or mixture of toxicants — compared with hundreds of thousands, or millions, of subjects that can easily be studied for height, body mass index, or major depression disorder).
Authors [see attached article] introduce two novel methods of the standard multivariate genome-wide-association meta-analysis (GWAMA) of related traits; these novel methods (N-GWAMA) and MA-GWAMA) are designed to correct for sample overlap. A broad range of simulation scenarios supported the added value of these novel multivariate methods, relative to the univariate GWAMA. Authors applied these novel methods to “life satisfaction, positive affect, neuroticism, and depressive symptoms” — which they refer to, collectively, as the “well-being spectrum.” The number of individuals in this GWAS is staggering: 2,370,390. Authors found 304 “significant independent signals.”
These new multivariate approaches resulted in a 26% increase in the number of independent signals — relative to the four univariate GWAMAs — and in an ~57% increase in the predictive power of polygenic risk scores. Supporting genome-wide transcriptomics, and methylomics, studies uncovered an additional 17, and 75, independent loci, respectively. Bioinformatics analyses, based on gene expression in brain tissues and cells, showed that genes differentially expressed in the subiculum and GABA-ergic inter-neurons (areas of the brain) are enriched — in their effect on “the well-being spectrum”. The (most recent) previous GEITP email described several GWAS, selecting the trait of “insomnia” to study, which some of us would consider very difficult to diagnose with any degree of quantification; however, next to “insomnia”, these GEITP pages would consider “the well-being spectrum” — even “softer” (i.e. more difficult to quantify). By adding these new bioinformatics and statistical methods (i.e. multlivariate vs univariate, in the case of this present article), these methods can be seen to boost the statistical power of finding associations between genotype and any “soft” phenotype. 🙂
Nat Genet Mar 2o19; 51: 445-451