PLINK 1.9-Heterozygous haploid warning-solution

[Prachi Chopra]

I tried plink 1.9 for solving heterozygous haploid warning using the following command-

'plink --file unclean_fileset --split-x b37 --make-bed --out clean_fileset'

Earlier I was getting 103317 heterozygous haploids but now it is decreased to 103205.

But still getting some errors which I have pasted below-

Warning: 103205 het. haploid genotypes present (see HbF_hh_clean.hh).
Warning: Nonmissing nonmale Y chromosome genotype(s) present.
Total genotyping rate is 0.996871.
894327 variants and 254 people pass filters and QC.
Error: --split-x cannot be used when the dataset already contains an XY region.
(Did you mean --merge-x instead?)

Then I used "--merge-x" for the warning "Nonmissing nonmale Y chromosome genotype(s) present"

It changed 646 chromosome codes and the heterozygous haploids are still same - 103205

Kindly help me out.

Thanks a lot

[Christopher Chang]

If an XY region is already present, you shouldn't use either --split-x or --merge-x here.

Do you have sex information for all samples?  If you don't, you'll need to use --check-sex/--impute-sex to fill in the blanks.

If heterozygous haploid calls still remain, the most likely cause is nonmissing female genotype calls on the Y chromosome; others have reported that this is fairly common.  A quick way to check the number of these is to just load the Y chromosome with e.g. "plink --bfile semi_clean_fileset --chr 24 --freq".  If all the heterozygous haploid errors are on the Y chromosome, you can safely clobber them with --make-bed + --set-hh-missing.  (If some are on the X, --set-hh-missing *might* still be okay, but I'd need to know more about the data source and the --check-sex report to be sure.)

[Vincent Laufer]

Chris - I think what would help me most is to understand the underlying issue about this error and why it exists. Is this described somewhere?

Thank you.

[Christopher Chang]

This is discussed under the "Warnings" section of https://www.cog-genomics.org/plink2/errors , but yes, it's hard to know to look there.  I will try to (finally) put together a good one-stop FAQ within the next month.

[kitty]

Hi Chris,

I wonder if you could help with this please.

I had both  the warning "1681151 het. haploid genotypes present" & "Nonmissing nonmale Y chromosome genotype(s) present" when I ran the --check-sex. Here are the codes I used:

plink --bfile file --chr 23-24 --make-bed --out myfile_unsplit
plink --bfile  myfile_unsplit --split-x b37 --make-bed --out split_myfile
plink --bfile split_myfile --indep-pairphase 20000 2000 0.5 --chr 23-24
plink --bfile split_myfile --extract plink.prune.in --make-bed split_myfile_pruned
plink --bfile split_myfile_pruned --check-sex

The output is very strange - I had ~2000 samples, but half of the sample had sex mismatch according to the output. Then I checked plink.hh file from the procedure you recommended as below ("plink --bfile semi_clean_fileset --chr 24 --freq"), it seems all the errors are on Y, so I clobber them with:
plink --bfile split_myfile_pruned --set-hh-missing --make-bed --out split_myfile_pruned_set-hh-missing

It improved the --check-sex results, those males wrongly assigned as "female" all got right, but still ~500 female samples are assigned as "male"(1) with large F ( > 0.97). I wonder, in your opinion, what the problem is likely to be please?

Many thanks,

KItty