Comparison of populations as a function of confidence intervals of gene probability

A new method for population comparison has been developed in order to determine the most discriminating systems in relation to diverse populations. The method compares pairs of populations on the basis of confidence intervals for allelic frequencies and does not require any previous assumption, such as normality of the population. A data set of 12 populations studied for Polymarker Kit (Roche) genetic systems, 13 biallelic systems (LDLR, GYPA, D7S8) and two systems with three alleles each (HBGG, GC), was considered. The analysis produced a total of 144 confidence intervals (12 populations x 12 alleles). The raw data were then analyzed comparing pairs of populations, allele by allele, through their respective confidence intervals. The method calculates the matrix of similarity between pairs of populations for each allele and eventually for each genetic system. Finally, the 12 populations were classified depending on the number of nondiscriminating alleles. The results of this method are compared to those produced by such usual techniques as Principal Component Analysis.