Population Structure Assessed Using Microsatellite and SNP Data: An Empirical Comparison in West African Cattle

Abstract

Simple SummaryProjection of genetic variability on geographic maps is a useful strategy to ascertain population structure and gene flow events when previous genetic information on the scenarios analyzed is not high. Here, we compared the performance of microsatellite sets and Single Nucleotide Polymorphism (SNP) arrays to identify the population structure and between-populations identity in a sample of West African cattle. Large SNP arrays were superior in detecting the population structure due to a more precise assessment of genotypic information of the individuals. However, the projection of genetic parameters on geographical maps was comparable between the SNP and microsatellite data. Geographic-based analyses of genetic variation areuseful inavoiding overinterpretation of the results obtained. Microsatellite markers can still be useful, particularly if the research focuses on non-model organisms or if either the funding or the availability of efficient hardware and software to handle large datasets is limited.A sample of 185 West African cattle belonging to nine different taurine, sanga, and zebu populations was typed using a set of 33 microsatellites and the BovineHD BeadChip of Illumina. The information provided by each type of marker was summarized via clustering methods and principal component analyses (PCA). The aim was to assess differences in performance between both marker types for the identification of population structure and the projection of genetic variability on geographical maps. In general, both microsatellites and Single Nucleotide Polymorphism (SNP) allowed us to differentiate taurine cattle from zebu and sanga cattle, which, in turn, would form a single population. Pearson and Spearman correlation coefficients computed among the admixture coefficients (fitting K = 2) and the eigenvectors corresponding to the first two factors identified using PCA on both microsatellite and SNP data were statistically significant (most of them having p < 0.0001) and high. However, SNP data allowed for a better fine-scale identification of population structure within taurine cattle: Lagunaire cattle from Benin were separated from two different N’Dama cattle samples. Furthermore, when clustering analyses assumed the existence of two parental populations only (K = 2), the SNPs could differentiate a different genetic background in Lagunaire and N’Dama cattle. Although the two N’Dama cattle populations had very different breeding histories, the microsatellite set could not separate the two N’Dama cattle populations. Classic bidimensional dispersion plots constructed using factors identified via PCA gave different shapes for microsatellites and SNPs: plots constructed using microsatellite polymorphism would suggest the existence of weakly differentiated, highly intermingled, subpopulations. However, the projection of the factors identified on synthetic maps gave comparable images. This would suggest that results on population structuring must be interpreted with caution. The geographic projection of genetic variation on synthetic maps avoids interpretations that go beyond the results obtained, particularly when previous information on the analyzed populations is scant. Factors influencing the performance of the projection of genetic parameters on geographic maps, together with restrictions that may affect the election of a given type of markers, are discussed.