Correcting wrong reference alleles

[marta-krych]

Hello Florian,

after importing PLINK files into R as a bigSNP object, I noticed that the reference alleles for all SNPs are inverted. To obtain the correct values, I should invert the 0s with the 2s within the genotype matrix, and additionally I should (probably) swap allele1 with allele2 for consistency. However, after trying different approaches, I haven't found a straightforward and memory-efficient way to handle this problem. Is there a way to effectively correct the encoding of the genotype matrix while ensuring that the NAs are preserved?

Thanks!

[privefl]

For consistency with what?

To switch 0s and 2s, you can simply use G2 <- G$copy(2 - G$code256); this won't make any deep copy, simply change the decoding.

[marta-krych]

I have previously tried to change the encoding in a similar way. However, when I attempt to do imputation on G2, I get the following error:

Error in snp_fastImputeSimple(G2, method = "mean0") : 
  identical(Gna$code256, CODE_012) is not TRUE

Should I consider G with the initial encoding for the imputation, and fix it afterwards? Or is there a way to switch 0s and 2s while keeping CODE_012 as the encoding for the matrix?

[privefl]

You can probably impute before changing the code.