search

rgcgithub / regenie

regenie is a C++ program for whole genome regression modelling of large genome-wide association studies.
https://rgcgithub.github.io/regenie
Other
189 stars 55 forks source link

pathway-based masks for rare variants association #462

Open igumnov-daniel opened 1 year ago

igumnov-daniel commented 1 year ago

Hi! I want to create the mask with variants from different chromosomes (pathway-based rare variants association). I've found in one of the closed issues the GENPOS in the variants set file doesn't affect the results and is only used to add the GENPOS in the result summary file. But what shall I do with the chromosome? As far as I understand for step 2 Regenie uses LOCO models from step 1. Thus the gene set I create must be located on one chromosome. Otherwise, Regenie's approach is incorrect (or at least doesn't address that variants are on different chromosomes anyhow). Am I right?

So is there a way to conduct pathway-based association with the current version of Regenie or is it just not possible due to Regenie's underlying model and logic?

Many thanks, Daniel

igumnov-daniel commented 1 year ago

Sorry, my bad. This topic was already discussed here: https://github.com/rgcgithub/regenie/issues/420

But I still do have a question: if I merge genotype files will Regenie work correctly despite the fact that in the variant set file I need to write chromosome №? Or Regenie doesn't care about both Chromosome and GENPOS in the variant set file?

Thank you

joellembatchou commented 1 year ago

Hi,

REGENIE uses chromosome value to select the step 1 LOCO predictor to use. Once sets contain variants across multiple chromosomes, it is not clear what chromosome to use... The consequence is that the step 1 LOCO predictor would capture some of the effects of variants being tested and hence lead to some power loss.

Cheers, Joelle