Slightly late into the fourth plenary: David Blair discussing "Common genetic variants associated with Mendelian disease severity revealed through cryptic phenotype analysis" #ASHG20
Cryptic phenotypes are phenotypes that underlie mendelian diseases, but which are not observed. For some, this will be a liability-threshold model (mendelian as extreme of normal range), but for others it will be a phenotypic outlier model (mendelian as truly separate) #ASHG20
Need models that differentiate between the models. May be morbidity-dependent genetic modifiers - e.g. may not see effects looking at the average of the population, but may see it at the extreme percentiles of severity #ASHG20
Estimated 50 cryptic phenotypes from EHR data, capturing symptoms of mendelian disorders. Estimated independent latent phenotypes, and generated two scores (spectrum versus outliers). Comparing the performance in a hold-out tests which model is better #ASHG20
Across 50 phenotypes, see a mix of results - many equivocal results, some show preference for one model over another. #ASHG20
Took spectrum phenotypes that replicated in UK Biobank forwards for common variant GWAS and G-E analysis. Examining the phenotypic extreme gives 1 locus, interquartile range -> 50 hits (many associated with population exposures like smoking). #ASHG20
Applied approach to Marfan syndrome. Extreme analysis identifies one robust GWS hit. Interquartile range analysis identifies three (not including the extreme analysis). #ASHG20
• • •
Missing some Tweet in this thread? You can try to
force a refresh
