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

its AAF in the ADD and ACAT model or it uses MAF ? #456

Open yasmeen-s opened 1 year ago

yasmeen-s commented 1 year ago

Hi @joellembatchou ,

Along with this issue https://github.com/rgcgithub/regenie/issues/256 and the other here: https://github.com/rgcgithub/regenie/issues/454 I have a similar question: If we specfiy via AAF file the variants we want to be tested, given in the AAF file we can only provide AF for ALT variants, like in PRS/GWAS we can flip alleles but here can we in regenie?.. Here in the gene burden testing situation when we specify --aaf-bins --build-mask max, and --joint acat

In the ADD model which is like a single varaint regression (but with multiple SNPs), in the output summary stats file what is 'BETA' here? It refers to which SNP/variant tested in a mask-aaf-bin? And second question is about the sign with 'BETA', how to interpret this 'BETA' if its + or - within various masks tested for a GENE, which are eventually combined in ACAT for a GENE . In the test as in equations its MAF (https://rgcgithub.github.io/regenie/overview/) but it will use 'AAF' from the AAF file if specfied as MAF ?

An example output for ACAT this doesn't assume same direction of effect but ADD model does ?

CHROM GENPOS ID ALLELE0 ALLELE1 A1FREQ N TEST BETA SE CHISQ LOG10P
22 17790634 MICAL3.M4b.5e-05 ref M4b.5e-05 0.00476388 180630 ADD 0.606864 0.227371 5.98837 1.84162
22 17790634 MICAL3.M4b.0.005 ref M4b.0.005 0.023598 180630 ADD -0.0885269 0.13941 0.403242 0.279494
22 17790634 MICAL3 NA NA NA 180640 ADD-BURDEN-ACAT 3.07692 1.10012 NA NA
joellembatchou commented 1 year ago

Hi,

(1) BETA refers to the effect of the burden mask obtained by aggregating all variants whose annotations belonged in the mask and whose AAF was below the mask AAF cutoff. (2) BURDEN-ACAT test does not have a corresponding effect size as it combines p-values of different burden masks using the ACAT method. Could you specify what version of REGENIE you are using in your example output above that has BETA/SE for BURDEN-ACAT?

Cheers, Joelle