Use of machine learning techniques for SNP based prediction of ancestry

Abstract

Some have argued that the genetic differences between continentally defined groups are relatively small and unlikely to have biomedical significance. In this study, the extent of variation between continentally defined groups was evaluated. Small numbers of randomly selected single nucleotide polymorphisms from the International HapMap Project were used to train classifiers for prediction of ancestral continent of origin. Predictive accuracy was then tested on independent data sets. A high degree of genetic similarity implies that groups will be difficult to distinguish, especially when only a limited amount of genetic information is used. It is shown that the genetic differences between continentally defined groups are sufficiently large that one can accurately predict ancestral continent of origin using only a minute, randomly selected fraction of the genetic variation present in the human genome. Genotype data from only 50 random single nucleotide polymorphisms can be used to predict ancestral continent of origin in the primary test data set with an average accuracy of 95%.

(cont.) Single nucleotide polymorphisms were also characterized as being in introns, coding exons, regulatory regions and regions coding for untranslated mRNA and classifiers constructed using only single nucleotide polymorphisms from a specific category. Predictive accuracy was similar across all of the classifiers created in this manner. Single nucleotide polymorphisms useful for prediction of ancestral continent of origin are common and distributed relatively evenly throughout the genome. These findings demonstrate the extent of variation between continentally defined groups and argue strongly against the contention that genetic differences between groups are too small to have biomedical significance.