Combining diploid+haploid data with admixture model

[CB]

Hi everyone,

This is more of a theoretical question than a methodological one--are
there any major problems with combining haploid plastid DNA data (or
mtDNA for that matter) with diploid nuclear data in a STRUCTURE
analysis under an admixture model?

I have run both a nuclear intron-only dataset and a combined nuclear-
plastid (haploid) data. In the latter case, I have coded the "other"
plastid copy as missing data. Am I introducing any bias into the
analysis by applying unjustified weight to the plastid data relative
to the nuclear data? If I understand this correctly, I am basically
forcing STRUCTURE to treat haploid data as if it were diploid. How
serious of a violation is this, and is there a way to circumvent it
while still combining haploid and diplod DNA?

Lastly, can anyone refer me to a study where this or something similar
has been done (e.g. mtDNA + microsats)?

Thanks!
CB

[Student]

Dear CB,

did you ever gain an answer to this question, I have the same.

Thanks Y.

[Jonathan Pritchard]

There should be no problem with combining ploidy in this way. For the
data that are haploid, the absent second copy should be coded as
missing data.

Note that for nonrecombining regions of the genome (eg mtDNA) the data
may violate the Structure models, and these data may be better coded
with each haplotype corresponding to a single allele.

Jonathan

[Student]

I apologize for the post-tennis!
So I have mtDNA and nuclear DNA.

Should I code both haplotypes/ alleles for each type of locus as a
single allele, so when asked for Ploidy in STRUCTURE = 1. Therefore,
for four nuclear loci and one mtDNA, I have 9 independent haplotypes/
alleles (columns).
Or
Should I code the second allele for mtDNA not as missing, but
duplicate the haplotype to make it diploid? Therefore, Ploidy in
STRUCTURE = 2, and four nuclear loci and one mtDNA, I have 10 columns
- five genotypes.

Also how would I know that assumptions were being violated if I did
include mtDNA? Or is that impossible to answer simply?

Thanks

[CB]

Dear Jonathan et al.,

Thanks very much for your replies, and I apologize for not writing
back sooner.

My nuclear data, under the delta K method, give nearly equivocal
results for K = 2 and K = 3, but when I include the plastid data, K is
unequivocally = 3. I have 3 a priori populations, so the K = 3
scenario makes the most biological sense to me, and argues for
including the plastid haploid data. My explanation is that my plastid
data show that there are three exclusive clades representing each of
these populations (i.e. reciprocal monophyly for each), and that my
nuclear data show either gene flow or lingering ancestral
polymorphism. This situation seems to be quite common at the
infraspecific level, and my guess is that it mostly has to do with the
difference in coalescence times for nuclear vs. plastid DNA. Thus it
is not surprising to me to see such results.

Can one specify different ploidy levels for different loci (i.e.
haploid, diploid) in a Structure analysis? How would this work in an
admixture analysis? I am hesitant to do what 'student' has suggested,
to simply code both alleles the same for haploid data (if I interpret
his or her meaning correctly). This is where I thought one would be
unjustifiably weighting the analysis (by doubling the influence of
haploid data, essentially treating them as a a completely homozygous
diploid locus), but I guess if one codes the other allele as missing,
the program will have to optimize that 'other' copy anyway, and the
same problem might still occur. For now I will stick with the latter.

Thanks again,

CB

> > > CB- Hide quoted text -
>
> - Show quoted text -