Chromosome X GWAS in related individuals
Entering edit mode
5.2 years ago
alesssia ▴ 580

Dear all,

I would like to include the X Chromosome in a GWAS of a continuous trait which has been measured in a dataset of related individuals (same-sex twins, ~22% males, 78% females).

I had a look at the literature, but most of the approaches presented seems to focus on case-control studies rather than on continuous traits. While some studies use PLINK, I am not sure it models X-inactivation correctly. Moreover, PLINK does not handle the presence of related individuals. Usually, we take inter-individual kinship into account by using LMMs, as implemented in GEMMA. However, I do not understand whether X-inactivation is or can be modelled in GEMMA.

Does anyone has any suggestion, or knows of any paper dealing with this?

Thank you very much in advance!

GWAS Chromosome X related individuals GEMMA • 1.3k views
Entering edit mode

PLINK's X-inactivation model is controlled by the --xchr-model flag. "--xchr-model 2" treats male genotypes as homozygous females, and is the default in PLINK 2.0.

With that said, PLINK doesn't currently have a LMM implementation.

Entering edit mode
5.2 years ago

Hey alessia, I am neither aware of any program that does this. I have heard that Clayton's method is used for modeling X-inactivation, but I am not sure in which programs Clayton's method is implemented. I am neither sure that PLINK can do what you need.

We had a somewhat similar issue in the past where we wanted to run GWAS over a large trios dataset and use interaction terms - there was nothing out there that could really do this. We eventually decided to do it ourselves in R, and I developed some parallelised code that could do the entire GWAS in a few days using ~64 CPU cores. This eventually became a Bioconductor package:

We exported data from PLINK and alleles were encoded as 1,2,3,4. I believe we used MAF tallies on continuous scale, and had other covariates like smoking status, cockroach exposure, and PC1/2 to control for population stratification. To control for family, we used conditional logistic regression with family as the matching stratum. So, model was something like:

clogit(Outcome ~ SNP:SmokingStatus + PC1 + PC2 + strata(family)

Not a direct answer, but my answer, nevertheless.



Login before adding your answer.

Traffic: 1368 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6