Clustering of the 312 tobacco accessions according to a model-based Bayesian algorithm implemented in the program STRUCTURE. Population memberships (expressed as%) for each accession are shown as estimates based on hypothetical subpopulations (see Methods). Each bar in the graph represents a single accession and its inferred proportion of admixture. The colors represent four different clusters corresponding to inferred unstructured subpopulations. The "Other" group includes genotypes of unknown type.