How to phase my own 23andMe data ?

Question

[LouisJoubert]

Hi, I am new this type of analysis but I am really interested in learning.

I did a DNA genotyping using 23andMe and I want to play a little bit with my data. My final goal would be to compare my personal data to population dataset, so I am taking it step by step.

So my question(s) are:

1) do I need to phase my data to perform what I want afterwards ?

2) if so how can I do this ? Using SHAPEIT or else ?

Thanks in advance to those taking the time to answer me.

Louis

Answer 1

[Charles Warden]

You'll need some additional data, like 1000 Genomes data.

While I am limited in my ability to provide assistance for the code, I have some examples of code to accomplish that here:

https://github.com/cwarden45/DTC_Scripts/tree/master/23andMe/Ancestry_plus_1000_Genomes

and here:

https://github.com/cwarden45/DTC_Scripts/tree/master/Genes_for_Good/RFMix_ReAnalysis

The RFMix portion is also based upon this code:

https://github.com/armartin/ancestry_pipeline

I would actually recommend Alicia Martin's code in terms of being easier to read (although I have some pointers on things that I was confused about in the Issues portion):

https://github.com/armartin/ancestry_pipeline/issues?utf8=%E2%9C%93&q=

[LouisJoubert]

Thanks, I will take a look at that !

[LouisJoubert]

Hi again,

How do you convert your raw 23andMe data to vcf ? Did you remove duplicates ?

What are the differences between the vcf.gz file you use in your scripts (ALL.chip.omni_broad_sanger_combined.20140818.snps.genotypes.vcf) and the files used in Alicia Martin's code (the same used in Kevin Blighe tutorial) ?

Thanks

[Charles Warden]

plink can accept the 23andMe file format (and there are probably other ways that others can perform the conversion). However, I essentially wrote some custom code (and excluded indels, where I didn't know the REF and VAR sequences).

So, I am not actually saying you should directly use my code. However, you should take time to understand all the steps to conversion, and I am OK with you getting some pointers from the code (although I would appreciate an acknowledgement, if you do that). From my end, I also realize that there is a limitation in what support I can provide, and therefore how much credit I can/should receive.

As for your question about which 1000 Genomes sample was used, I selected the Omni array since the uncompressed version was much smaller than an Illumina-sequencing-based multi-sample .vcf.

I don't think Alicia added any new samples, so she didn't have to worry about combining files in different formats (which is unfortunately directly related to your question - however, the combination of code may help with creating a combined file that is compatible with downstream analysis).