I am currently working on a STRUCTURE analysis of 45 populations of pedunculate oak in Europe, consisting of 1120 individuals, based on genotypic data from 21 microsatellite loci. Given wind-pollination, fragmentation may have only weak effect on genetic structure. On the other hand, plantations are commonly used for regeneration and it is almost certain that some of the populations (or individuals) are of foreign origin.

Im my STRUCTURE runs I set K=1...20, admixture model, correlated allele frequencies, 10 runs for each K.

Evanno's ΔΚ provides some weak evidence for K=16. For K=16, ΔΚ is maximized (however its value is only 6) and variation of ln probability of data is obviously lower. On the other hand, ln probability of data is highest for K=1 and decreases for higher values of K, which speaks against structure (in presence of structure, it increases up to the most likely K and then "plateaus"). Partition as seen from q-values is not very obvious, but can be seen with some imagination. Here are the files: https://drive.google.com/folderview?id=0B41pt4LUC4K3dEE5ZUF6WkU1Y2M&usp=sharing

What is your opinion? K=1 or K=16?

I tend to conclude that there is no partition. However, I have seen some similar examples in the literature (e.g. http://link.springer.com/article/10.1007/s11295-009-0216-y#page-1), being interpreted as evidence about subdivision. In addition, I have reasons to believe that there are some oak stands significantly differentiated from the others (e.g. significant Nei's distances from all other pops, "foreign" chloroplast DNA). On the other hand, STRUCTURE could miss a genetically differentiated group if this is represented by relatively few individuals.

In the light of these remarks/thoughts and given STRUCTURE results, is it maybe an alternative to mention that there is weak evidence for population subdivision?

More Charalambos Neophytou's questions See All
Similar questions and discussions