Did you get an answer for this? I have the same problem

Hi Cansu,

Thank you for this answer. I will check my bim file so.

I will let you all know if mine worked.

Thanks a lot

Hi,

you're probably over with your analyses. I am writing how I solved this problem for the people in the future.

So, I started with a .vcf file and converted it to the custom plink format. So, treemix requires a specifically formatted file. The website says that in order to create this file you need to run:

$ plink --bfile data --freq --missing --within data.clust

$ gzip plink.frq

$ plink2treemix.py plink.frq.gz treemix.frq.gz

The first codeline ($1) requires a .clst file. .clst files are generated with --write-cluster --family and it consists in three columns including I. Family ID; 2. ID; 3. cluster. I used several awk commands to format the information properly.

The first codeline ($1) produces a .frq file which according to plink2's webiste includes:

I. CHR [Chromosome code]

II. SNP [Variant identifier]

III. A1 [Allele 1 (usually minor)]

IV. A2 [Allele 2 (usually major)]

V. MAF [Allele 1 frequency]

VI. NCHROBS [Number of allele observations]

So, here start the problems. If you open the provided python file:

$ nano plink2treemix.py

You will notice that the script requires that the file has 8 columns (remember that python starts counting in 0. "mc = line[6]", "total = line[7]" are columns 7 and 8 respectively). I think this might be a problem with the plink versions. I was using plink2 (v1.9) and the program was written when plink1 (v1.07) was around. So if you hit with the following error:

ERROR: treemix mc = line[6] indexerror: list index out of range

I expect it to be a mismatch between plinks versions.

So then I switched to plink1 and it complained that the SNPs had similar names. Because my data was generated with STACKS denovo, the SNPs do not have aposition in the genome and had "." for SNP. I used awk to solve this:

$ awk '{$2 = $1"_"$4; print}' mydata.bim > mydata.SNPrenamed.bim # so, we change the second column ($2) to be a combination including column 1, an underscore and column 4 ( $1"_"$4 ).

Plink kept complaining so I decided to take a new approach. I noticed that the POULATIONS mode as part of STACKS 1 ( http://catchenlab.life.illinois.edu/stacks/manual-v1/ ) had a treemix flag.

I created a population map for stacks, which is a file composed of two columns, the first one with individuals and the second with populations:

So I just went on STACKS v1.48 and did:

/path/to//stacks-1.48/bin/populations --in_vcf mydata.vcf --treemix -O ./ -M pop_map.tsv

--in_vcf : input vcf

--treemix : generate a treemix output

-O : output folder

-M : population map

And it worked!

Hi José,

Thank you for this answer. I hope too that it will help researchers doing such analyses and get it done faster.

vcftools --gzvcf DK.vcf.gz --plink-tped --out DK

plink --tfile DK --freq --noweb --missing --within pop.cov

Have you resloved this questions? When I went on STACKS v1.48 and did:

/path/to//stacks-1.48/bin/populations --in_vcf mydata.vcf --treemix -O ./ -M pop_map.tsv, this was not working, and produced only several summarize data that was not related to treemix.

The newest version of Stacks (2.3) supports a TreeMix output option in the populations command. I have recently used this and no modifications of the file are needed to go directly into the TreeMix software.