Merging inconsistencies in PLINK

Question

[GabrielMontenegro]

I am trying to merge two different PLINK files genotyped on the same platform. However, I found some inconsistencies that could not be solved by the --flip command. I checked the SNPs that could not be merged after the flip and found this type of problem:

File 1:

4    rs10000432    68.93    47511781    T    C

File 2:

4    rs10000432    68.93    47511781    A    C

One of the alleles is different, while the other allele is the same on both files.

What should I do on this case? I wasn't expecting to find these type of issues on these datasets since both were genotyped on the same platform.

[Kevin Blighe]

Just an initial comment: for that particular SNP, the ancestral 'reference' allele is C (https://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=10000432).

When you attempted to merge the first time, PLINK may have output a file with the extension *.missnp, which would contain these multi-allelic sites. You can remove these from your dataset with the following command:

plink --noweb --bfile MyData1 --exclude MyData.missnp --make-bed --out MyData1.Pruned ;
plink --noweb --bfile MyData2 --exclude MyData.missnp --make-bed --out MyData2.Pruned ;

Then attempt to merge again.

If you don't want to necessarily remove these, then you may have to do more rigorous data preparation. In which format was your data initially - VCF?? It would be useful to run every genotype against a reference genome and ensure that the ref>alt order is maintained.