I'm studying the genetic bases of phenology traits collected in two consecutive years using FarmCPU; to detect significant associations I relied on an FDR of 5%. My sample size is 162 individuals for one year and 161 for the following one; the SNP panels differ by 1,000 between years, and the inter-year distributions of a given trait differ from each other.
I have found several significant SNPs bearing the same effect associate to the traits in both years: in one year, consecutive, same-effect SNPs can map as far as 100k bp from each other, and some groups of SNPs with the same effect can map on different chromosomes. Moreover, there are groups of SNPs with almost mirror effect (e.g. +/- 137) to other ones. The traits measured in the other year have way less signals, and those SNPs with the same effect are very close to each other.
All the SNPs bearing the same effect are in linkage disequilibrium with each other and are found in the same genotype proportions in the population; some of these SNPs violate HW equilibrium. The mirror effect were, instead, due to either the minor or the major allele being used as a reference (ancestral status has not been resolved yet for my species).
Running the GWAS without the SNPs the violate HW proportions, I still retrieve signals bearing the same effect and (far less numerous than before) mirror effect. Some of the latter ones do map far from each other, but this time they alway remain circumscribed to along the same chromosome.
It escapes me how FarmCPU returns as significant SNPs that are in LD, given that it should control for it even across chromosomes. Has anyone some insight about it? Thanks in advance.
