search

yoonsucho / CAMERA

Mendelian Randomization approach using Cross Ancestral Model
Other
6 stars 5 forks source link

Heterogeneity of exposures as indication of difference in instrument effects #11

Open explodecomputer opened 2 years ago

explodecomputer commented 2 years ago

Slope could tell us general difference but if there is no heterogeneity then that implies all the effects are relatively constant

To choose SNPs that are zero-relevant, two approaches

  1. SNPs that have high power for replication but don't replicate
  2. SNPs that have high heterogeneity but don't replicate

Pleiotropy of the SNP is just the effect in pop2 where gx=0.

Is there a way to aggregate this across all zero-relevance SNPs?

When calculating alpha 1 and alpha 2, adjust bgy1 and bgy2, see if it reduces heterogeneity

yoonsucho commented 2 years ago

[brain storming...]

Obtain heterogeneity Q for pop1 exposure on pop2 exposure: Using 2SMR package -> instrument_heterogeneity() function is added

Get per-SNP q value : Using RadialMR package -> test_pleiotropy() function is added

SNPs that have high power for replication but don't replicate: A list of the SNPs can be obtained from estimate_instrument_specificity() SNPs that have high heterogeneity but don't replicate: Obtained by using Radial MR (v) =======Done=======

To investigate whether those putative pleiotropic SNPs are pleiotropic indeed