beagle 4 exception: Duplicate marker

Question

[galbarel]

I'm trying to use BEAGLE 4 to do an IBD analysis.

I'm running the command: java -jar beagle.r1398.jar gt=vcf_format.vcf out=beagle ibd=true

where vcf_format.vcf is a file I generated using PLINK 1.9 (with --recode vcf-iid) from my PLINK format files (bed, bim, fam).

BEAGLE stops with an exception:

Duplicate marker: 1    72765116    rs2568958_r    A    G

I've searched for this marker in both the VCF file and the BIM file, and it appears only once in both.

Can anyone suggest what to do? Or how can I check differently for duplicates in my files?

Thanks

Answer 1

[chrchang523]

BEAGLE 4's marker equality check is based on chromosome, position, alleles, and the INFO field's END key if one is present; it does not consider marker ID. (For the curious, this is in src/vcf/Marker.java lines 473-509.) So you probably have another marker in your file with the same chrom/pos/alleles, and these could very well be genuine duplicates that you want to resolve with e.g. PLINK --exclude.

I will add a --detect-duplicate-var flag to PLINK today or tomorrow to help out here.

[chrchang523]

Implemented as --list-duplicate-vars in the December 9 development build. Sample workflow, if you don't have multiple variants with the same ID:

plink --bfile my_dataset --list-duplicate-vars ids-only suppress-first --out tmp
plink --bfile my_dataset --exclude tmp.dupvar --recode vcf-iid --out vcf_format
java -jar beagle.r1398.jar gt=vcf_format.vcf ...

[Amir]

Thank you chrchang523.

This is R script ready to run.

# Set working directory
setwd("YOUR WORKING DIRECTORY")

# Function to run plink
runPLINK <- function(PLINKoptions = "") system(paste("plink directory/plink.exe", PLINKoptions))
runPLINK()

dir.create("results")

runPLINK("--vcf VCF_file.vcf --list-duplicate-vars ids-only suppress-first --allow-extra-chr --out results/tmp --memory 16000")
runPLINK("--vcf VCF_file.vcf --exclude results/tmp.dupvar --recode vcf-iid --allow-extra-chr --out results/vcf_format")

Answer 2

[Jie Ping]

I am not an expert on this. But I have read a paper whick did an IBD analysis using PLINK directly.

I am confused that why not do an IBD analysis using PLINK directly?

[galbarel]

You are correct, there is an option to do IBD analysis with PLINK, but I have been suggested to use BEAGLE's fastIBD algorithm since it is faster and might have better results.