Why different reference panels change the signal for a SNP?

Question

[anamaria]

Hello,

can someone please explain to me why here for fig 2 https://www-nature-com.proxy.cc.uic.edu/articles/ng.3643/figures/2

signal for this SNP changes between different versions of reference panels. In particular in between HapMap 2 and HRC?

Here is a description of those two panels:https://imputationserver.readthedocs.io/en/latest/reference-panels/

Answer 1

[curious]

I can't see your paper, its behind a paywall but is it just showing better signal in association from imputing with HRC?

Take a look at that second link what are the big differences between hapmap and HRC?

Also take a look at the benefits section of the HRC website, might give you some clues:

http://www.haplotype-reference-consortium.org/

[anamaria]

Thanks for the link. Yes I understood that the power to detect the low frequency SNPs increases...HRC panel has much more SNPs and is done on immensely more subjects. I also did a plot to compare my imputed data with HRC ("new") with imputed data for HapMap2 ("old") and indeed that is the case. Also only around 850000 SNPs are overlapping between imputation with those two reference panels. So new significant variants can be found with HRC. But I am wondering if GWAS for data imputed with HapMap 2 showed a signal for a particular SNP is it possible that with imputation with HRC that signal is lost (the p value for that SNP is much higher)

Bdw the paper name is: "A reference panel of 64,976 haplotypes for genotype imputation" and I refer to Fig2.