Our key takeaways can be found in these longer threads. https://twitter.com/arbelharpak/status/1125825650386927617 …https://twitter.com/molly_przew/status/1126150189612400640 …
-
-
Prikaži ovu nit
-
One analysis that we had not highlighted previously asks why SNP heritability estimates vary dramatically across (socio-economic status/sex/age) strata.pic.twitter.com/CPiJx4hF9n
Prikaži ovu nit -
We initially expected heritability differences to have little to do with genetic effects; and that they are mostly due to independent differences in environmental variance across strata. However, such a model does not fit the data.pic.twitter.com/SzWkmyFAEQ
Prikaži ovu nit -
We hypothesize instead that there are large differences in genetic variance across strata. In other words, this is potential evidence for substantial GxE.
Prikaži ovu nit -
Thanks to a reviewer suggestion, we matched the “diverse” GWAS (eg a sample of an equal number of men and women) sample size to the non-diverse GWAS (eg only men or only women). This helps demonstrate that the performance/portability of diverse samples is not necessarily better.pic.twitter.com/cAhZFRMD7k
Prikaži ovu nit -
We developed a test to detect the substantial effect of either population structure confounding, assortative mating or indirect genetic effects. But there is still work ahead to *quantify* the contribution of direct genetic effects vs. other factors to polygenic scores.pic.twitter.com/tLYJ1U1hND
Prikaži ovu nit -
One more lingering question: can the portability problem be mitigated by borrowing ideas from other fields, e.g. where the learning and prediction sets are known to differ in important ways? (<-- not a rhetorical question) https://twitter.com/anshulkundaje/status/1125887976896995329 … h/t
@anshulkundajePrikaži ovu nit
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.