Assessing spatial genetic structure from molecular marker data via principal component analyses: A case study in a Prosopis sp. forest

Advances in genotyping technology, such as molecular markers, have noticeably improved our capacity to characterize genomes at multiple loci. Concomitantly, the methodological framework to analyze genetic data has expanded, and keeping abreast with the latest statistical developments to analyze molecular marker data in the context of spatial genetics has become a difficult task. Most methods in spatial statistics are devoted to univariate data whereas the nature of molecular marker data is highly dimensional. Multivariate methods are aimed at finding proximities between entities characterized by multiple variables by summarizing information in few synthetic variables. In particular, Principal Component analysis (PCA) has been used to study genetic structure of geo-referenced allele frequency profiles, incorporating spatial information with a posteriori analysis. Conversely, the recently developed spatially restricted PCA (sPCA) explicitly includes spatial data in the optimization criterion. In this work, we compared the results of the application of PCA and sPCA in the study of the spatial genetic structure at fine scale of a Prosopis flexuosa and P. chilensis hybrid swarm. Data consisted in the genetic characterization of 87 trees sampled in Córdoba, Argentina and genotyped at six microsatellites, which yielded 72 alleles. As expected, principal components explained more variance than sPCA components, but were less spatially autocorrelated. The maps obtained by the interpolation of sPC1 values allowed a better visualization of a patchy spatial pattern of genetic variability than the PC1 synthetic map. We also proposed a PC-sPC scatter plot of allele loadings to better understand the allele contributions to spatial genetic variability.