SNP-Trait Association

Question

I have a population ~4k individuals genotyped with ~40k SNPs. My phenotype is a binary trait. Most individuals have score 0 and only nine individuals have score 1, and 41 missing values. I am trying to fit also 6 covariates as random effects by using SNP-Trait Association option in JMP Genomics. I am getting the following error message.

ERROR: PROC GLIMMIX generated no estimates for the Genotype Test for trait Shredded. There may be excessive missing data or the contrast may be nonestimable. Please see the log for more details.

Does the algorithm allow for missing values?

Thanks

Chris_Kirchberg · Answer

HI @Fers ,

I would contact support@jmp.com at this point and they should be able to help troubleshoot why you are still getting an error.  Please let them know you are using JMP Genomics, which version and your site number so they can make sure you get to the correct person.

Sorry I could not be more help.

Fers · Answer

The genotypic data is already imputed, so missing is not an issue. But I have aroud 45 samples with missing phenotype. I already tried by removing the samples with missing values and it seems I have other troubles.

ERROR: PROC GLIMMIX generated no estimates for the Genotype Test for trait Shr

Thanks

Chris_Kirchberg · Answer

@Fers,

Another thought is that if those markers and samples need to be there, you can impute the genotype using nearest neighbors. In JMP Genomics 9 this process was added and can be found under Genetics Utilities. This might be helpful given the proportion of missing genotypes is small relative to the total number of genotypes.  Also to consider, are these missing values due to array or sequencing errors? Or do they hold some special meaning (not enought information, to variable to call, etc.)?