Finally, if “admixture LD” is present, so that in admixed individuals long segments of the genome originate from one founder population, simple PCA methods will not be as powerful as programs such as STRUCTURE [10], ADMIXMAP [11], and ANCESTRYMAP [12], where there is careful modeling of the admixture blocks and the transitions. The power of these methods lies in the fact that genome-wide samples may have similar proportions of inheritance from the ancestral populations, but locally they will inherit either 0, 1, or 2 alleles from each ancestral population. Methods that specifically attempt to assign local ancestries will be able to determine the specific patterns typical of each ancestral population locally. An interesting and challenging problem is to build tools that retain the power of these more complex models on admixed data and that also run rapidly on large datasets.
