samtools Option to get genotype likelihood (GL)

[hansdd]

I would like samtools mpileup to generate genotype likelihoods or what is described in the VCF FORMAT as GL. I can get PL which is fine (I know how to convert it to probabilities), but vcftools requires GL to generate the likelihood input files for Beagle.

Which option in samtools/bcftools generates the GL info per animal and genotype? Apologies if this should be obvious...

Cheers,

Hans


Here is the command I am using now:
samtools-0.1.18

samtools mpileup -r Chr29:46186560-48558888 -P ILLUMINA -guf all_UMD3.fa $files | bcftools view -bcvg - > var.test.bcf

bcftools view var.test.bcf > var.test.vcf

[Heisman]

These work for me:

/samtools-0.1.18/samtools mpileup -l <intervals_to_call_variant_on> -AB -ugf <reference_sequence.fa> <aligned_file> | /samtools-0.1.18/bcftools/bcftools view -bvcg -> <intermediate_file>


/samtools-0.1.18/bcftools/bcftools view <intermediate_file> | /samtools-0.1.18/bcftools/vcfutils.pl varFilter -D99999 > <output> &

[hansdd]

Thank you for the response. I tried the suggested commands but samtools still doesn't list the GL field.

Is it a discontinued option?

[Heisman]

Sorry, I get the PL output, not the GL (PL is the phred scaled versions of GL, I think). I will play around with this and report back...

Edit: Yeah, I can't figure it out. Maybe email the author about this?

[cjp]

I think this is about right:

PL ~ int(-GL * 10)

So:

GL ~ -PL/10

As PL is given as an integer, you'll lose some precision to GL, but both samtools and GATK seem to give PL now.

Chris

[cjp]

Also: vcf2gen_v1.1.pl will calculate genotype probs from GL and PL - the script is here:

http://www.stats.ox.ac.uk/~jsliu/software/

Chris

[Heisman]

Also: vcf2gen_v1.1.pl will calculate genotype probs from GL and PL - the script is here:

http://www.stats.ox.ac.uk/~jsliu/software/

Chris

Interesting, I was under the impression that since the PL is scaled such that the lowest value is 0, it wouldn't be possible to revert back to GL without knowing the scaling factor?

[cjp]

GL can be 0.00 as well. PL=0 just means the Prob. is very high and effectively 1 (or at least heading towards 1).

[hansdd]

Thank you for the help.