How does GT output in imputed vcf correspond to DS, HDS and GP? chromosome X imputation Michigan server

[tesselgalesloot]

Hi all,

A few years ago (2017), I have imputed chromosome X genotype data using the Michigan imputation server. I discovered the following in the output field GT:DS:GP:

0/0:0.705:0.403,0.489,0.108
DS and GP would result in most likely GT 0/1 (DS closest to 1, GP highest prob for the heterozygote), whereas GT is 0/0.

I checked by imputing the same file again, and now the default output also contains field HDS. I now think this phased DS output is used for GT determination, is this correct? See for example:

output fields: GT:DS:HDS:GP

• 0|0:0.695:0.212,0.483:0.407,0.490,0.102 --> I would expect heterozygote based on DS (0.695 closer to 1 than 0) and GP (0.490 is highest prob), but based on HDS, 0|0 is likely, as both dosages are below 0.5.
• 1|1:1.404:0.502,0.902:0.049,0.499,0.453 --> I would expect heterozygote based on DS and GP, but 1|1 is logical based on HDS (2 times > 0.5).

I would really like to know if my reasoning is correct or whether there is another reason for the 'discrepancy'.

Thank you in advance for your reactions!
Best,
Tessel