Is Association Testing With Mpileup A Sound Approach?

Question

[Travis]

Hi all,

I'm not a stats person so forgive if the question seems basic.

I know it is possible to perform an association test with samtools mpileup, but I am wondering how this compares to using dedicated softwares like PLINK for example? I have no hand-on experience of association testing.

I guess I am wondering if a bioinformatics scientist (for example) should feel happy doing his own association test with samtools or if he should be handing the SNP data off to the statistical department for more rigorous studying?

I guess there's a line here but I'm not sure at what point it exists...

Thanks in advance.

Answer 1

[Larry_Parnell]

While the question is basic, it is highly relevant. We are a genetics lab and that means we look very rigorously at the statistical methods used to determine association between allele(s) and phenotype(s). We don't use PLINK because we feel that GAPIT is superior, especially as family structure can be important. The methods behind GAPIT are outlined here and a paper specifically on this is in preparation.

In addition, if the association tests are a component of a manuscript, you can expect to receive Reviewer comments pertaining to the statistical methods used and tests undertaken. Hence, the recommendation to be rigorous. Spurious statistical findings are no friend of the scientist.

[Larry_Parnell]

No, I cannot offer any comments as Samtools and mPileup are not part of our analysis pipelines - we don't generate sequence data, e.g. We would advocate putting the data into PLINK or GAPIT or something designed for association tests. Again, if there is any type of family relationships in your data, this is vital to control for. I don't expect that Samtools can account for this. Also to consider: controling for sex, age, disease status, etc.

[kumar.vinod81]

The only problem with GAPIT that it treats all the missing values in genotype file as heterozygotes?

[Travis]

Thanks for the comments Larry. Do you have any particular comments regarding the shortcomings of Samtools mPileup for performing association tests?

[Travis]

Thanks a lot Larry.