accuracy of imputed genotypes

[felassadi]

Hello Dr. Simone,

I used GLIMPSE1 for imputation. My data was sequenced at low coverage (1X), but we sequenced one sample at high coverage (30X) to run concordance and check the accuracy of imputation. I have 114 samples, and I ran glimpse-phase after merging the genotype likelihoods of all samples. I used a modified version of 1000k+HGDP haplotypes.
In an older issue, you responded to someone else saying that concordance is under testing and we can't be confident about its results. My first question is that still the case? The concordance plots I generated showed r2 <0.8 for MAF<2%
My second question is related to filtering out poorly imputed genotypes. For that, I used "bcftools filter -i 'INFO>0.8' ", the output filtered file was significantly smaller in size than the original file (which was generated by glimpse-ligate). This means a lot of the imputed genotypes have INFO<0.8. For downstream analysis, can I use the non-filtered file, or I should proceed with the filtered file?
Is there any advice you can give me?

Thank you in advance.