Adding genetic relationship matrix as covariate in SKAT

[nimisha chaturvedi]

Hello Shawn,

I am using SKAT for running a omnibus test on haplotypes and I have two questions regarding the analysis:

1. I would like to correct it for genetic relationship between samples. For that I use the GRM matrix from GCTA (this is a square matrix with values in the lower triangle of the matrix = values in the upper triangle of the matrix and diagonal values =1) and try to create a null model (response ~ GRM matrix) using SKAT_NULL_Model. The problem is that when I do this nul$s2 becomes NaN and then I cannot proceed with running the analysis using SKAT anymore. I am wondering if this is happening because of the structural property of the matrix (lower triangle = upper triangle) and if so then is there a way to resolve the issue in SKAT? 

2. I use amino acid residues as independent dataset for running the analysis, which are coded as 0,1,2. The issue is that this 0,1,2 coding does not takes into account MAF. So when I use the data with SKAT, it starts flipping the "genotypes" because it is designed for rare variants. Is there a way to stop this in SKAT? 

Thanks!

Nimisha

[Seunggeun (Shawn) Lee]

Hi Nimisha,

1. If you want to adjust for the Kinship, you need to use SKAT_NULL_emmaX instead of SKAT_NULL_Model. In SKAT_NULL_emmaX, you can provide a GRM matrix as a parameter (K).

2. If Z is not a genotype, please use is_check_genotype=FALSE in SKAT function. 

Thanks,

Shawn

[Alessandro Gialluisi]

Hi all,
I am trying to input a kinship matrix in SKAT_NULL_emmaX(), but I haven´t found so far a format that the package likes.

What I have tried so far is:

1. a GRM matrix produced by GCTA, as mentioned here (https://groups.google.com/forum/#!searchin/skat_slee/SKAT_NULL_emmaX$20kinship$20matrix%7Csort:relevance/skat_slee/jpohFKJ2lVQ/esCc2tmPEAAJ). Of course, I read the matrix through the R function proposed on the GCTA official page (http://cnsgenomics.com/software/gcta/estimate_grm.html). In this case, I get the errror message  K is not a matrix!
2. a kinship file as produced by Emmax (through epacts make-gin command). In this case, I specify the Kin.file field instead of K field in SKAT_NULL_emmaX(), which apparently should work anyway in absence of a kinship matrix. However, i get the following error message

            Read  GENO.kinship
           Error in scan(file = file, what = what, sep = sep, quote = quote, dec = dec,  :
           scan() richiede 'a real', invece risulta 'EMMA_KIN`'

Maybe you know how I can produce a kinship matrix in a format tha SKAT likes? Hope some you can give a feedback on this...

Thanks a lot in advance.

Best,

Alessandro

[Alessandro Gialluisi]

Hi all,
just to give an update: a kinship file produced through Emmax make-kin seems to work..however, I am still curious to know which software produces a matrix in a format that SKAT likes..so any feedback from people with experience on this is welcome!!

Best,
Alessandro