Post by glenn on May 18, 2016 15:42:52 GMT
Hi,
I am attempting to estimate the heritability of quantitative trait (body mass) for a wild bird species. The dataset I am using is 175 individuals collected across the species range and genotyped at 5154 biallelic loci. These loci were generated from a RADseq run. This is a non-model system so I do not have a reference genome, therefore no reliable information on the genomic positions of these SNPs.
Following the GCTA tutorial, I have estimated the genetic relationship matrix (GRM) using the command. There were no errors or warnings.
>gcta_mac --bfile test --autosome --maf 0.01 --make-grm --out test
When I run the following command to estimate the variance in body mass explained by the SNP, I get V(G)/Vp = 0.999995. So something isn't quite right.
>gcta64 --grm test --pheno test.phen --reml --out test
There is strong population stratification in this species, so perhaps that's the problem. I included the first 10 PC axes of the SNP matrix as covariates. When I run the command
>gcta64 --grm test --pheno test.phen --reml --out test
I get the following error...
Error: analysis stopped because more than half of the variance components are constrained. The result would be unreliable.
Please have a try to add the option --reml-no-constrain.
When I run the code with no constraint
>gcta64 --grm test --pheno test.phen --reml-no-constrain --out test
I get the following error...
Error: the matrix V becomes negative-definite because of one of the variance component is negative.
Please re-run the analysis without the --reml-no-constrain option.
Either way, GCTA does not like it when I include the genetic PC axes as covariates. Does anyone know what might be the problem? Do I have big enough sample sizes to even be attempting this analysis?
Many thanks,
Glenn
I am attempting to estimate the heritability of quantitative trait (body mass) for a wild bird species. The dataset I am using is 175 individuals collected across the species range and genotyped at 5154 biallelic loci. These loci were generated from a RADseq run. This is a non-model system so I do not have a reference genome, therefore no reliable information on the genomic positions of these SNPs.
Following the GCTA tutorial, I have estimated the genetic relationship matrix (GRM) using the command. There were no errors or warnings.
>gcta_mac --bfile test --autosome --maf 0.01 --make-grm --out test
When I run the following command to estimate the variance in body mass explained by the SNP, I get V(G)/Vp = 0.999995. So something isn't quite right.
>gcta64 --grm test --pheno test.phen --reml --out test
There is strong population stratification in this species, so perhaps that's the problem. I included the first 10 PC axes of the SNP matrix as covariates. When I run the command
>gcta64 --grm test --pheno test.phen --reml --out test
I get the following error...
Error: analysis stopped because more than half of the variance components are constrained. The result would be unreliable.
Please have a try to add the option --reml-no-constrain.
When I run the code with no constraint
>gcta64 --grm test --pheno test.phen --reml-no-constrain --out test
I get the following error...
Error: the matrix V becomes negative-definite because of one of the variance component is negative.
Please re-run the analysis without the --reml-no-constrain option.
Either way, GCTA does not like it when I include the genetic PC axes as covariates. Does anyone know what might be the problem? Do I have big enough sample sizes to even be attempting this analysis?
Many thanks,
Glenn