Post date: Dec 06, 2013 8:50:45 PM
I would like to know whether the genetic basis of performance variation predicts genetic differences between GLA and SLA. In particular, are alleles that increase fitness on Ms more abundant in GLA than SLA. Addressing this question is tricky. First, we have two different estimates of the genetic architecture of performance on Ms, one for GLA and one for SLA (both seem relevant). Second individuals SNPs are weakly associated with performance at best. Third somewhat different sets of SNPs were included in each mapping exercise as I used SNPs with maf > 1% in the relevant population in each case. Finally, I use performance associated SNPs from the combined analysis of SLA and GLA but the need to correct for differences in survival (taking us off of the binary scale) and genetic structure make this less than ideal (and perhaps not worthwhile). My first approach was to consider the SNPs with the highest PIP for Ms survival in GLA and SLA or GLA only (those with PIP > 0.01, which is still pretty low and only includes 22 SNPs even in the combined case), and ask whether more of these had the expected allele frequency difference (reference in GLA > SLA if beta is positive and GLA < SLA if beta is negative) than expected by chance. This does not appear to be the case, but I am not sure this means much given the genetic architecture (many possibly or weakly associated SNPs). Next I want to try estimating a breeding value for each wild individual based on the sparse effect estimates only (i.e., beta * gamma; the truly polygenic component cannot be used given the relatedness structure). I would do this using the effect estimates for GLA and SLA on Ms but make comparisons only within a set. My prediction is that the mean breeding value of GLA individuals will be higher than that of SLA individuals based on either or both sets of effect estimates.