Post by bill on Jan 6, 2015 15:46:11 GMT
Hello,
I was wondering if GCTA would allow me to test gene associations with a continuous phenotype, but using a continuous summary statistic for each gene instead of raw genotypes. In other words, rather than encoding genotype as 0,1,2 (number of minor alleles), I will be aggregating SNPs within a gene and computing a summary statistic for the gene that reflects the expected impact that the SNPs will jointly have on the protein encoded by the gene (i.e., using SIFT or Polyphen, etc.). This is similar to a burden test, but with an arbitrary summary statistic summarizing the joint effect of all SNPs in the gene and their predicted effect on protein structure/function. I'd like to use this summary statistic when testing for an association between the gene and a continuous phenotype:
Y ~ BX+u +e
where Y is a continuous phenotype measure, B is effect, X is the summary statistic from aggregating SNPs in a gene and applying SIFT/Polyphen, u is the random effect controlling for sample structure (population stratification, cryptic relatedness, etc.), and e is error. Thus, X is an arbitrary continuous value, rather than the usual 0,1,2 encoding.
Is this possible in GCTA?
Sorry if this has already been asked here.
Thanks so much,
Bill
I was wondering if GCTA would allow me to test gene associations with a continuous phenotype, but using a continuous summary statistic for each gene instead of raw genotypes. In other words, rather than encoding genotype as 0,1,2 (number of minor alleles), I will be aggregating SNPs within a gene and computing a summary statistic for the gene that reflects the expected impact that the SNPs will jointly have on the protein encoded by the gene (i.e., using SIFT or Polyphen, etc.). This is similar to a burden test, but with an arbitrary summary statistic summarizing the joint effect of all SNPs in the gene and their predicted effect on protein structure/function. I'd like to use this summary statistic when testing for an association between the gene and a continuous phenotype:
Y ~ BX+u +e
where Y is a continuous phenotype measure, B is effect, X is the summary statistic from aggregating SNPs in a gene and applying SIFT/Polyphen, u is the random effect controlling for sample structure (population stratification, cryptic relatedness, etc.), and e is error. Thus, X is an arbitrary continuous value, rather than the usual 0,1,2 encoding.
Is this possible in GCTA?
Sorry if this has already been asked here.
Thanks so much,
Bill
