Why mach1 is not imputing SNPs?

Question

[moxu]

The command lines for the two steps:

mach1 -p mysamples.22.merlin.ped -d mysamples.dat --vcfReference -h ~/xdata/reference/1kg/b37/ALL.chr22.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz --compact -r 30 --prefix mysamples.mach1step1

mach1 -p mysamples.22.merlin.ped -d mysamples.22.dat --vcfReference -h ~/xdata/reference/1kg/b37/ALL.chr22.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz  --crossover mysamples.22.mach1step1.rec --errormap  mysamples.22.mach1step1.erate --greedy --geno --quality --dosage --probs --mle --mldetails  --autoFlip  --mask 0.05 --prefix mysamples.22.mach1step2

The problem:

I have two 6758 SNPs in mysamples.22.merlin.ped & mysamples.dat as the input files for step 1. But after step 2, the result files (.mlinfo, .mlgeno, .mldose) still contain 6758 SNPs. No more, no less.

The question:

Why did the imputed SNPs go?

Thanks!

Answer 1

[moxu]

I think it's because mach1 does not take the option "--vcfReference".

mach-admix takes "--vcfReference" but the 1000 Genome phase3 reference files contain duplicated SNPs (e.g. ALL.chr22.phase3_shapeit2_mvncall_integrated_v5a.20130502.genotypes.vcf.gz has two entries for rs7410429) and mach-admix aborts when it sees duplicated SNPs.

What do you do for the duplicated SNPs?