How to bypass SNPs with identical A1 and A2 alleles in PLINK?

Question

[yorgos.athanasiadis]

Hello,

I am trying to merge two imputed (I used SHAPEIT and IMPUTE2) binary file sets with PLINK's bmerge command, but this error pops up:

Error: Identical A1 and A2 alleles on line 1

I am pretty sure I've got many single-allele SNPs in my data, so I was wondering if there is a quick way to solve this problem? I checked PLINK's manual, but there seems to be no way to ignore such SNPs or correct them.

I would like to avoid - if possible - to look for a solution in SHAPEIT or IMPUTE2, because prephasing and imputation already took a very long time to run.

Any ideas?

[chrchang523]

This might be due to an incompatibility between PLINK's Oxford import and the latest IMPUTE2 output format. Can you send me a small .gen/.sample fileset that generates this problem? (You can probably omit all lines in the .gen file past the first 3-4.)

Answer 1

[yorgos.athanasiadis]

Sure:

Gen (impute2) file (these are actually the first lines of the file - SNPs in the converted plink file have a different order):

--- 1:55565:G:A 55565 G A 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1
--- 1:55582:T:C 55582 T C 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1
--- 1:55588:T:C 55588 T C 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1

Sample file:

ID_1 ID_2 missing sex 1
0 0 0 D 1
_0005432396f6cb97_FAM _0005432396f6cb97 0 2 1
_0022055d7384cd08_FAM _0022055d7384cd08 0 2 1
_0121e9eb137f1c02_FAM _0121e9eb137f1c02 0 1 1
_0187a92841fa96a2_FAM _0187a92841fa96a2 0 2 1
_018e122ad9751cf7_FAM _018e122ad9751cf7 0 2 1

I hope this helps

Yorgos

[chrchang523]

Did any of these SNPs end up with equal A1/A2 alleles? (I just tried importing and merging them, and did not have any problems.)

• If only other SNPs were affected, could you find the corresponding .gen line for one of them?
• If you actually did have problems with these specific SNPs, can you list the commands you used to import and merge your files?

[yorgos.athanasiadis]

Hi again,

No, I didn't have any problem with these specific SNPs, they just were the first ones in the impute2 file and I tried to do exactly what you asked me to (i.e. paste the 3-4 first lines). I guess that SNP order was changed when I converted from gen to plink format.

I have spotted the line in the impute2 file where the monomorphic SNP is:

1 rs12564807 734462 A A 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0 0 1 0

• All the lines I've pasted here are incomplete - they are actually quite longer, involving many more individuals
• I have used regex with grep + awk commands to create a list of such monomorphic SNPs and then --exclude them with PLINK. This actually resolves the issue, I was just wondering if PLINK has some built-in flag to ignore this problem, which by the way pops up only when I try to use the --merge commands.

[chrchang523]

There currently isn't a built-in command, since this generally indicates a data processing error that should be fixed at the source. But do you know what caused impute2 to generate a file with identical A1 and A2 allele codes here? If it's a routine occurrence, and it only happens with monomorphic SNPs, I will modify the .gen (and .bgen, if necessary) import routines to automatically zero out one of the allele codes here.

[yorgos.athanasiadis]

Hi,

After looking back at all the files, I found the following:

1. the original bim file I used in PLINK to prephase with SHAPEIT had this line for this specific SNP:

1    rs12564807    0    734462    0    A

2. Using that file, SHAPEIT returned this haplotype for the SNP (Showing only part of it here):

1 rs12564807 734462 A A 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

I guess that that '0' (instead of 'G') is responsible for the issue.

Y.