STRUCTURE is a commonly used program to detect population structure among humans using mutilocu genotype data. The program was originally developed to control for population structure for association studies, but this program and other structure-like programs are widely used for genetic studies of human to understand human prehistory and population subdivision patterns.
STRUCTURE is a model based program, which means that the results obtained are only an approximation under the assumptions that model is build on. Therefore, the applicability of the program has to be considered carefully by examining its assumptions.
This method assumes that Hardy-Weingberg equilibrium and linkage equilibrium. There should be no evolutionary force acting on the populations being analyzed (no mutation, no natural selection, no genetic drift, random mating, and population size should be large), except for gene flow which are included in the model to estimate admixture proportion of individuals. The genetic markers chosen have to be unlinked, non-gene coding markers.
However, the human populations violate these assumptions. Mating among human are not random and non-gene coding markers could be linked to allele are naturally selected. The populations size of hunter-gatherers is usually very small, so genetic drift have affected their genetic variations. The genetic markers used for analysis have to be selected carefully, because the long linkage disequilibrium is observed in small populations and recently admixed populations.
I do not think Pritchard et al (2000) address some other possibly important issues. How the changes in population size affect the program? What if ancestral populations did not evolve independently? What happen, if one of the ancestral populations was genetically diverse and other did not have much diversity?
Maybe, is newer version of STRUCTURE more suitable for human population genetic studies?
Tidak ada komentar:
Posting Komentar