The Russian gene pool: Gene geography of surnames

Surnames are traditionally used in population genetics as "quasi-genetic" m arkers (i.e., analogs of genes) when studying the structure of the gene poo l and the factors of its microevolution. In this study, spatial variation o f Russian surnames was analyzed with the use of computer-based gene geograp hy. Gene geography of surnames was demonstrated to be promising for populat ion studies on the total Russian gene pool. Frequencies of surnames were st udied in 64 sel'sovets (rural communities; a total of 33 thousand persons) of 52 raions (districts) of 22 oblasts (regions) of the European part of Ru ssia. For each of 75 widespread surnames, an electronic map of its frequenc y was constructed. Summary maps of principal components were drawn based on all maps of individual surnames. The first 5 of 75 principal components ac counted for half of the total variance, which indicates high resolving powe r of surnames. The map of the first principal component exhibits a trend di rected from the northwestern to the eastern regions of the area studied. Th e trend of the second component was directed from the southwestern to the n orthern regions of the area studied, i.e., it was close to latitudinal. Thi s trend almost coincided with the latitudinal trend of principal components for three sets of data (genetic, anthropological, and dermatoglyphical). T herefore, the latitudinal trend may be considered the main direction of var iation of the Russian gene pool. The similarity between the main scenarios for the genetic and quasi-genetic markers demonstrates the effectiveness of the use of surnames for analysis of the Russian gene pool. In view of the dispute between R. Sokal and L.L. Cavalli-Sforza about the effects of false correlations, the maps of principal components of Russian surnames were co nstructed by two methods: through analysis of maps and through direct analy sis of original data on the frequencies of surnames. An almost complete coi ncidence of these maps (correlation coefficient rho = 0.96) indicates that, taking into account the reliability of the data, the resultant maps of pri ncipal components have no errors of false correlations.