Haplotype frequencies

[mrw3288]

I am interested in looking at some very short regions (~200 bp) in the human genome which contain ~15 SNPs. So, taking for instance Phase 3 phased haplotype data (vcf) from 1000 genomes (~2500 individuals), I would like to identify all the different haplotypes and count them, thus allowing me to obtain the haplotype frequencies in this sample of individuals.

I have tried using 'vcfgeno2haplo' and 'vcf2tsv' which are part of vcflib on Galaxy but I cannot the first to accept my data; can someone suggest how I might do this or where I should look?

There's an interesting tool described at

http://www.biomedcentral.com/1471-2105/15/200

for visualizing this information but it does not enable one to pull out the statistical data that I need.

Thanks.

[cowman]

I am trying to answer the same question but on a genome scale. What are the haplotype frequencies at each locus across all loci.
Plink --blocks will generate lists of haplotype blocks but there is only one block per locus which implies a maximum of two haplotypes per locus in a population which cannot always be true.
Impute and Shapeit will phase genotype data but I have not found any sets of haplotype blocks assoicated with them so even if their models contemplate more than two haplotypes at a locus it is not possible to estract this information from the phased data.
Any ideas about how I should find the number and frequencies of haploytpes at each locus in the 1000 genomes data?

[mrw3288]

A colleague of mine should supply the code. But I don't understand what you mean by "at each locus across all loci"; what is your "locus" or what sort of "marker" are you considering?

[cowman]

Hi Mrw3288, thanks for your reply.
I mean SNP loci. I will be working with 1000 genomes data. I want to know the number of haplotypes at each SNP locus in the genome in each population in the data set.
However locus can be a flexible concept and I could work with windows of say 10kb.

[mrw3288]

So next what do you mean by "at each SNP locus"? You need at least two (adjacent) SNPs (along a chromosome) to start making a haplotype.
Anyway, if I solve my question (200 bp window) I should be able to solve yours (10 kb window), and I'll let you know if/when that happens.

[cowman]

True, but haplotypes have boundaries at SNP so the number of haplotypes associated with one SNP may be different from the adjacent SNP. The resolution that I will work to will partly depend on speed and practicality.
I will look forward to hearing of your solution.

[mrw3288]

Here (attached) is a partial solution; it's an R script written by my colleague Jaqueline Wang. Ignore the comments in Portuguese. Put the script and your *.vcf.gz (eg sliced from 1000genomes) file in the same directory and run the script. The output is a *.tsv file. You'll have to slice the vcf according to your needs.

[cowman]

Thanks a lot for that. I will let you know how I get on.

[Sofia T]

Here (attached) is a partial solution; it's an R script written by my colleague Jaqueline Wang. Ignore the comments in Portuguese. Put the script and your *.vcf.gz (eg sliced from 1000genomes) file in the same directory and run the script. The output is a *.tsv file. You'll have to slice the vcf according to your needs.

Hi! I would be very interested in this script, but it is not available for download now. Would you be able to share it again? Thank you!

[mrw3288]

Hi! I would be very interested in this script, but it is not available for download now. Would you be able to share it again? Thank you!

Hi. The project is finished. Try contacting Jaqueline Wang using [email protected] who may be able to provide you with the script.