Tajima's D

[Craig Anderson]

Heya,

I’m trying to calculate Tajima’s D across the genome using GBS data that’s gone through Stacks and so far have tried to calculate it per site using pi (from Stacks), ensuring that sample size changes with respect to missing data at each site. However, using standard calculations for theta, a1, a2, e1, e2 and with S=1 at every site, my D scores are very high across the genome, with 4’s and 5’s occurring frequently.

 I’ve looked at other bits of software, such as vcftools and popgenome, but they associate sample size more broadly with the number of individuals total, so genome-wide data become difficult to correct when results reflect sliding windows. I’m aware that such stats have taken a knock (http://onlinelibrary.wiley.com/doi/10.1111/mec.12276/full), but I’m still keen to try. Is it possible to emulate the Hohenlohe et al. 2010 paper using the current Stacks output, so as to determine sample size within the sliding window without performing the analysis independently of Stacks? Is it even worth it?

Cheers,

Craig

[Janna Fierst]

Hi,

I have found that pi is completely coverage-dependent in whole genome sequencing data, and sequencing coverage is very uneven and biased by local GC content and other quirks. I haven't calculated pi for our RAD data but our allele frequency estimates show that if anything, RAD is far more biased than the whole genome data. We met with Eric Johnson's group about these issues and the problem is that the null expectation for sequencing data is not that the number of reads would reflect allele frequencies. The null expectation varies by locus with GC content, restriction bias, and sequencing bias and we dont know what it is- so methods that assume a certain null are problematic. We use Fisher's exact for our whole genome data, could you use other assumption-free measures to get at your questions? -Janna

--
Stacks website: http://creskolab.uoregon.edu/stacks/
---
You received this message because you are subscribed to the Google Groups "Stacks" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stacks-users...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.