STRUCTURE assumption HWE and polymorphism

[Élise Mazé]

Hi everyone,
I have a question about how STRUCTURE deals with data from introduced species. That is, species which are not in HWE, and sometimes with a lack of polymorphism at some locis (e.g. 2-3 alleles for 1 loci). Is STRUCTURE subject to pitfalls with these conditions? I don't find publications on this subject...Have someone any idea?
Many thanks in advance,

Mazé Elise

[Vikram Chhatre]

Elise,

Generally speaking, STRUCTURE assumes that loci are at HWE and LE (linkage equilibrium) within populations.  You will therefore need to use caution in interpreting results depending upon how much these loci are deviating from such assumptions.  You may want to consult the user guide and Pritchard et al (2000) page#946, column#2, paragraph#2 for more information on modeling assumptions in STRUCTURE.

Out of curiosity, what introduced species are you working with and what type of markers?

Also, 2-3 alleles at a given locus is at worst low polymorphism, not lack of polymorphism.  SNPs are by definition, biallelic.  

Vikram

--
You received this message because you are subscribed to the Google Groups "structure-software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to structure-softw...@googlegroups.com.
To post to this group, send email to structure...@googlegroups.com.
Visit this group at http://groups.google.com/group/structure-software.
For more options, visit https://groups.google.com/groups/opt_out.

[Julie Hebert]

Elise,

I would just add that almost every program that uses allele frequencies to analyze genetic data assumes HWE at some point, yet as scientists, we know HWE is very seldom met. I think one of the keys to studies that use programs such as Structure is to make sure you have large enough sample sizes to get good estimates of given populations, and enough markers to accurately assess the variation present. I have primarily worked with AFLPs and although I exclude singletons, I typically have some loci where only two individuals have the minority allele. The key is that I have a lot of other loci that have different frequencies of the minority allele to balance the other alleles out. (Does that make sense?) So if these are the only loci you have, then I would definitely question the results, but if these are just a few loci relative to your total genetic data, any problems relating to the lack of polymorphism should come out in the wash.

You can also do complementary analyses such as NMDS to look to see if similar patterns emerge. There are several statistical packages that will do NMDS. I usually use vegan in R.

Julie

On Saturday, December 7, 2013 9:43:00 AM UTC-5, Vikram Chhatre wrote:

Elise,

Generally speaking, STRUCTURE assumes that loci are at HWE and LE (linkage equilibrium) within populations.  You will therefore need to use caution in interpreting results depending upon how much these loci are deviating from such assumptions.  You may want to consult the user guide and Pritchard et al (2000) page#946, column#2, paragraph#2 for more information on modeling assumptions in STRUCTURE.

Out of curiosity, what introduced species are you working with and what type of markers?

Also, 2-3 alleles at a given locus is at worst low polymorphism, not lack of polymorphism.  SNPs are by definition, biallelic.  

Vikram

On Thu, Dec 5, 2013 at 3:38 AM, Élise Mazé <neli...@gmail.com> wrote:

Hi everyone,
I have a question about how STRUCTURE deals with data from introduced species. That is, species which are not in HWE, and sometimes with a lack of polymorphism at some locis (e.g. 2-3 alleles for 1 loci). Is STRUCTURE subject to pitfalls with these conditions? I don't find publications on this subject...Have someone any idea?
Many thanks in advance,

Mazé Elise

--
You received this message because you are subscribed to the Google Groups "structure-software" group.

To unsubscribe from this group and stop receiving emails from it, send an email to structure-software+unsub...@googlegroups.com.

[Adii_ (Institute of Genetics and Animal Breeding)]

Hi Julie. I just want to say: R rocks! ;)

Adii_

[Aaron]

Hi, if your data is out of HWE and/or has linkage disequilibrium, you might want to look into using sPCA using the adegenet package in R. From my reading of the paper associated with it (Jombart et al. 2010 I think), HWE and LE are not assumptions of the method.
-Aaron