Question

assocTestSingle output with GxE

0

Entering edit mode

SuperDad • 0

@6c37bad3

Last seen 6 months ago

United States

Hi there,

I know the assocTestSingle can output effect size estimates for the genotype (G) term and interaction (GXE) term when GXE is applied. Is there any way to get the effect size estimate for E term too?

Thanks

GENESIS • 764 views

ADD COMMENT • link updated 10 months ago by Stephanie M. Gogarten ▴ 870 • written 10 months ago by SuperDad • 0

score 1 · Answer 1 · 2023-06-21

1

Entering edit mode

Stephanie M. Gogarten ▴ 870

@stephanie-m-gogarten-5121

Last seen 4 weeks ago

University of Washington

You can find the effect size estimate for the E term in the output of fitNullModel, in the fixef list element.

ADD COMMENT • link 10 months ago Stephanie M. Gogarten ▴ 870

0

Entering edit mode

Is there any way to involve G and GXE as null model, and then test E effect estimate and its p-value.

such as

null model= covariates + G + GXE

test model= covariates + G + GXE +E

or

null model= covariates + G

test model= covariates + G + GXE +E

Thanks

ADD REPLY • link 10 months ago SuperDad • 0

0

Entering edit mode

You could just use the fitNullModel function and add a genotype vector to the input data frame. Then you can look at all the effect sizes in the same place. There's no difference in the results between doing that and running assocTestSingle; the reason we separate the null model from the model with genotypes is for computational efficiency (the null model only needs to be computed once, then we can test the genotypes in parallel).

ADD REPLY • link 10 months ago Stephanie M. Gogarten ▴ 870

0

Entering edit mode

Thanks. Can I add GxE=c("E") in the fitNullModel? I would like to use the full model covariates + G + GXE +E in the fitNullModel.

ADD REPLY • link 10 months ago SuperDad • 0

1

Entering edit mode

Yes, but the syntax is different. You specify an interaction with ":":

annot <- data.frame(outcome=rnorm(10), G=sample(c(0,1,2),10,replace=TRUE), E=rnorm(10))
mod <- fitNullModel(annot, outcome="outcome", covars=c("G", "G:E", "E"))
mod$fixef
               Estimate Std. Error     t value   Pr(>|t|)
(Intercept)  0.37250092  0.6177681  0.60297859 0.56859873
G           -0.02477967  0.3884110 -0.06379755 0.95120379
E            1.85738947  0.7968132  2.33102248 0.05855874
G:E         -1.48079165  0.4919364 -3.01012853 0.02369659

ADD REPLY • link 10 months ago Stephanie M. Gogarten ▴ 870