rad loci - how many?

[Sara Rocha]

Dear Peter, all,

I am new to migrate - will use for first time. I think I reasonably understand what it does but have no experience in how long it takes to run ...

I've read the 2019 current protocols in bioinformatics papers and made the tutorials and I am going through the manual before I start. Also went though the threads in this group for some insights. 

The goal is to estimate Ne and migration rates between three (geographic) populations of fish... divergence (if exists) is very very recent and gf high... from previous admixture analyses the 3 pops are not easily distinguishable.... 

I was planning to use a SNP dataset subsampled (we have ~400K snps coming from rad experiment) but from what I read it is not the best data to use... so I am now thinking of trying with (a subsample of) the full rad loci (we have around 400 loci of between 100 and 500 bp). 

I have some questions before I start, hoping someone can give me some help:

1) in case I want to try the SNPs, I see the vcf script mentioned in the biobootcamp tutorial is not distributed is version 5. Any reason not to use it (appart from analysing snps being not the best idea)?...

2) If using the radloci... how much is too much for current migrate version? should say 50 loci do the job or will it already be a hard one for migrate? I can use the mpi version in a large cluster.

Thanks in advance for any insight,

best,

sara

[Sara Rocha]

one xtra question I forgot: from STACKS one gets haplotype guesses (=2 seqs per ind / locus). These can be inputed directly into migrate given different "individual" labels, right? and in this case the nr of individuals = nr of sequences, right (i.e = individualsX2) ?

thanks in advance again

sara

[Peter Beerli]

Dear Sara,

On May 22, 2024, at 7:27 AM, Sara Rocha <spr...@gmail.com> wrote:

The goal is to estimate Ne and migration rates between three (geographic) populations of fish... divergence (if exists) is very very recent and gf high... from previous admixture analyses the 3 pops are not easily distinguishable.... 

I was planning to use a SNP dataset subsampled (we have ~400K snps coming from rad experiment) but from what I read it is not the best data to use... so I am now thinking of trying with (a subsample of) the full rad loci (we have around 400 loci of between 100 and 500 bp). 

this is the far better approach than using snps.

I have some questions before I start, hoping someone can give me some help:

1) in case I want to try the SNPs, I see the vcf script mentioned in the biobootcamp tutorial is not distributed is version 5. Any reason not to use it (appart from analysing snps being not the best idea)?…

I will need to check on that because I thought that it is in the contribution folder.

The problem with these type of codes is often that I have not tested them extensively, but if you have vcf files then try them, here a link to the code and other data converters that I used over time [no warranty that they work but they give a good start in case someone want to improve on them]: https://github.com/pbeerli/dataconverters.git

2) If using the radloci... how much is too much for current migrate version? should say 50 loci do the job or will it already be a hard one for migrate? I can use the mpi version in a large cluster.

using many loci (say 1000 or so) should not be a problem, if you use the parallel version, the single cpu version will take a long time.

This question comes up relatively frequently, I should construct a tutorial for that.

best

peter

Thanks in advance for any insight,

best,

sara

--
You received this message because you are subscribed to the Google Groups "migrate-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to migrate-suppo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/migrate-support/bd4968d1-b1b4-4e8a-99df-9dc6c4bc9838n%40googlegroups.com.

[Sara Rocha]

Took me a while to get back to this.

Thanks a lot Peter!!