Help with allele effect interpretation

[David Moseley]

Please help me understand the allele effect values a little more. I have attached an excel file that shows the alleles I found associated with my trait (seed length). From Tassel I have included the major allele and minor allele.

For every allele, I notice Tassel shows one has an allele effect of 0 and the other one has an actual value (either positive or negative). I have included the allele that tassel had a value which is under the SNP affecting column. In the next column I have the allele effect from Tassel.

In the next column, I have a allele effect I calculated using Excel. I did this by averaging the values from the major allele and subtracting from the average value of the minor allele. In this case I always had the value showing the effect of the major allele. (Major-Minor).

I see that if Tassel showed the minor allele with a value, my allele effect (always Major-Minor)  has the inverse sign. (- or +). This at least makes sense, however, I do not know why Tassel has one with a value and the other with a 0. Is it not true that if the minor allele has a negative effect, then the Major allele has a positive effect?

My values are sometimes the same number, but sometimes are bigger or smaller. Does Tassel factor in more steps than just (Avg. Major-Avg. Minor)?

I also show the allelic effect from GAPIT which seems to always be the inverse from the Tassel value and mostly about half the value. What does GAPIT do different?

Please verify these concerns, if you would.

Thanks,

David

[Peter Bradbury]

In TASSEL, the allele effects are actually genotype effects. For example A would be AA and C would be CC, so the effect of an allele substitution would be (AA - CC) / 2 or (CC -  AA) /2. The sign is arbitrary. The actual values calculated for the genotype effects is dependent on how the linear model is formulated. If there are two genotypes, say AA and CC, there is only a single degree of freedom available to test their difference or to estimate a genotype effect, so you cannot estimate genotype effects without making an additional assumption. TASSEL does this by setting one of the genotypes to zero. In a sense, the genotype effect estimates themselves are not meaningful (non-estimable) but their difference is. Making direct comparisons between TASSEL and GAPIT is tricky because they use somewhat different linear models. TASSEL fits an additive + dominance model while GAPIT fits an additive only model.

Peter

[David Moseley]

Thank you. Do you have an article I can use in my paper to reference this?

Thanks again. 

[abhinandan patil]

Great reply 

I doubt cleared with this answer 

many thanks

Peter

[David González]

Dear Peter,

I got confuse about the model tassel uses to estimate the marker effect. From the Wiki, it explains that MLM model additive random effects (kinship) by coding the marker genotype as 2,1,0, So relationship are based on the additive substitution effect. But for the marker fixed effect, it doesn't indicates how the marker is coded.. (2,1,0)??? to estimate the dominance effect of makers, I guess it should include another column (covariate) for each marker coded as e.g. (0,1,0), for the heterozygote in order to estimate the Dominance effect. 

In the "stat report", are the statistical effects (i.e. substitution effects and dominance deviations) reported? 

Please, Could you clarify how TASSEL model both additive and dominant effects. Thanks in advance!

Kind regards,

David