Hi,
I'm trying to phase (using Eagle) and impute (using minimac4) some variant data using the Michigan Imputation Server from 1233 Japanese individuals. I have autosome vcfs for each chromosome (except Y, as these are female individuals), but for each file I try to add I get very low reference overlap:
For the 1000G -phase3 reference panels (for hg38), the server only gives the "ALL" and the "mixed" reference options (I tried the "ALL", but the reference overlap was much lower, and QC was failed completely)
These are the only settings where QC was passed, so I'm not sure what I'm doing incorrectly for the reference coverage to be so low.
I am new to this sort of stuff, so I'm wondering if there might be a step prior to this that I have missed and might need to do before submitting to this server.
•
written
by
graeme.thorn
How did you prepare your input data for imputation? Did you perform HRC or 1000G Pre-imputation Checks? Pre-imputation check is very essential before you submit your data to Michigan Imputation Server. See the link below for how you can perform it-
https://imputationserver.readthedocs.io/en/latest/prepare-your-data/
Additionally, please make sure that you choose correct build of your data (hg19 or hg38).
Thanks. These are hg38 samples, so I'd need an hg38 reference panel for the explicit checks.
The previous data I was given was all hg19 and used MIS with minimac3 for imputation, but this new data I've got seem to have major issues with being lifted over (to hg19) to use the same method (for consistency purposes)
You do not have to lift over. There is reference for hg38 too for pre-imputaion check. See in the link below- https://www.well.ox.ac.uk/~wrayner/tools/
I would also suggest you to try Asian reference panel since your data contains Japanese subjects. You will find info in the same page in the link above.
Log in to answer this question.
