I am using a 2-step approach to imputation: phasing with hapiur, followed by imputation with minimac. For a given snp SNP1, I am interested in what influenced the dosages output by minimac. In the reference panel, there are 3 snps in complete LD (r2=1) with SNP1 in the assumed ancestry group for my data. The LD is slightly less than perfect when phase is considered, and there is 1 genotype that does not correspond when looking outside the ancestry group. Other than these 3 snps, no other snps within 30kb have r2>.14 with SNP1. One of the snps (SNP2) in LD with SNP1 (r2=1) was directly typed in my data. However, if I take a "Best Guess" approach to interpreting the imputed dosages, by rounding them and converting to genotypes using an additive model, there are quite a few genotypes that do not correlate between SNP1 and SNP2. I am wondering if there is an easy way to tell what variants/haplotypes influenced dosage for SNP1.
I should note that the overall RSQR for SNP1 output by minimac is .97, which (although possibly coincidentally) is the same as the r2 between SNP1 and SNP2 if phasing is taken into account.
