Genotype imputation without a reference panel

Question

[Mohadese]

I have genotyping array data, for some of the individuals we have missing data even on the genotyping array. We want to impute those missing data based on the rest of the dataset. What tools can I use?

[karl.stamm]

Be aware the individual may have good reason to not get a call at that site. Maybe the gene is deleted or duplicated and has no clear diploid genotype. Imputation will put in a likely result, but not the sure result.

Answer 1

[Pine]

I might be wrong but I am not sure if using the same genotyping data internally would work in a classic imputation setting. My impression is that you need a separate reference panel for imputation beside the genotype data you want to analyze. If lack of reference panel is an issue, some online servers provide preloaded reference panels, for instance:

Michigan imputation server: https://imputationserver.sph.umich.edu/index.html#!

Guidelines here: https://imputationserver.readthedocs.io/en/latest/

I suspect you might be able to fill in the missing dosage values with population average for instance (how the missing values are sometimes treated for non-genetic variables) using packages like SNPStat in R which give you full access to genotype matrices. But I am not personally familiar with studies that have used this approach for genotyping data.

[Mohadese]

Unfortunately we are not able to use online servers like Michigan or TOPMed.

[Pine]

Many genotype reference panels and imputation packages are publicly available [including the ones used by Michigan imputation server] and can be downloaded for a local imputation. My understanding is that local imputation is also a common practice, but unfortunately I do not have personal experience with it.