Dear all,

I would like to perform an eQTLs analysis with OpenMx using a IBD matrix inferred by using genotype data. The model I would like to evaluate is the ACQE one. Does someone have some scripts I can use/adapt or some hints about how to deal with this?

Thank you very much!

Alessia

Hi Alessia,

It's possible I'm in the minority here, but I know I had to do a fair amount of searching just to make headway understanding the question. For anyone who is similarly ignorant, eQTLs stands for expression quantitative trait loci (eQTLs), IBD stands for identity-by-descent (IBD), and I still haven't found out what ACQE means. What's ACQE?

I don't know of anyone that has scripts for this type of model. Perhaps relatedly, OpenMx is very close to a public release of its version of genone-wide complex trait analysis (a more general extension of what is possible with the GCTA software). This is the work of Rob Kirkpatrick.

HTH!

Hi, Alessia. Since Mike Hunter mentioned me in this thread already, I'll reply to your post. I hesitated to reply previously, because I'm not sure how helpful I can be.

I'm pretty sure I understand your question, but not completely, so let me make sure. It sounds like you want to do some kind of linkage analysis, where the "phenotype" is expression level of a gene. You already have genotype data from your participants on a panel of markers, and since you can infer non-trivial IBD amongst your participants, you must have closely-related individuals in your sample. Further, you're interested in ACQE, meaning that you can identify a model including variance components for additive-genetic variance, shared-environmental variance, variance due to IBD at a given locus, and non-shared environmental variance; therefore, you probably have twins in your dataset. Is this correct?

Assuming I understand correctly...I unfortunately don't know of any OpenMx scripts for the purpose, but I have an idea of how they would work. In addition to the usual A, C, and E variance components, there would also be a Q component; total phenotypic variance would be A + C + Q + E. The covariance between two relatives would be due to C, some proportion of A, and pihat * Q, where pihat is the (inferred) proportion of alleles shared IBD at the locus in question. You would include the IBD proportion in the model as a "definition variable." I guess your script would have to loop over all of the loci of interest and re-fit the MxModel at each one.

Starting on page 145 of the manual for "classic" Mx, there's an example script you could maybe try to adapt to OpenMx. You might also consult: