Melanesian gene frequencies: a multivariate data-analytic approach

Abstract

Arcsine transformations of gene frequencies for the ABO, Rh, MNSs and Gm systems, derived from published sources for New Guinea and other Western Pacific populations were subjected to multifactorial analysis using a modified Mahalanobis d 2 . The plot of the difference between the computed distances from the overall multivariate mean against the quantiles of the gamma distribution revealed a departure from linearity. Two Polynesian Outlier Islands, Bellona and Ontong Java, contribute to this distorted distribution, as also do Ulawa and Biak. A cluster analysis of the genetic distances between the 134 populations studied indicates a general cohesion in Island Melanesia, with similarities between populations of the Solomons, New Hebrides, New Caledonia, Fiji and Austronesian (AN) New Guinea populations. Some non-Austronesian (NAN) New Guinea groups also fall close to this cluster. Highland populations in New Guinea form another cluster, with some groups from the Sepik and Markham Rivers intruding, supporting other evidence of gene flow between the highlands and these two valleys. Discriminant analysis of AN and NAN groups gave a high percentage of correct classification and the implications of this are discussed. Finally, a principal component analysis based on transformed Gm frequencies, is used to examine the distinctiveness of AN and NAN groups. It is concluded that using this set of markers it is futile to seek a distinction between AN and NAN populations which is universal. It is probable that there have been complicated local patterns of admixture, not a simple dichotomy.