Theobroma cacao L., or cacao, is the source of cocoa products used in the making of chocolate. These tropical trees are conserved in living genebanks. The International Cocoa Genebank, Trinidad is one of the largest ex situ collections in the public domain. Mislabelling is a critical problem and the correction of this problem is vital to improve the accuracy and efficiency of genebank management. Using microsatellite DNA markers, we assessed the level of mislabelling in a group of Refractario cacao that originated from Ecuador and determined their population memberships through Bayesian clustering analysis. The microsatellite data revealed a synonymous rate of 7.5% and an error rate of 39.4% in this germplasm subset. The analysis of the population structure grouped the Refractario accessions into four subclusters, indicating intra-population heterogeneity in this germplasm group. Based on the results, we recommend that when the assignment test is used for cacao genotype identification, it should (a) use duplicated samples as internal checks, (b) choose suitable reference accessions, including a known homogeneous group and (c) employ subclustering checks to obtain reliable results. The information framework generated is discussed in relation to the management of the collection, population enhancement and future research of the collection.

The International Cocoa Genebank, Trinidad, is the largest field genebank collection of cacao (Theobroma cacao L.) in the public domain and the correct identity of each tree is crucial for germplasm movement, evaluation and phenotypic characterization. Nine microsatellite loci were used to assess the identity of 1477 trees from 486 cacao accessions representing approximately 16.9% of the trees and 29.2% of the accessions within the genebank. Heterogeneous plots (plots containing more than one genotype group) averaged 25.1% in The International Cocoa Genebank, Trinidad, with maximal admixture (32.6%) being recorded in Field 5B. The error rate did not differ significantly among different fields. Mislabeling error could be affected by accession grouping with an average error rate of 27.4% for accession groups in the genebank. Synonymous accessions were estimated to account for 14.4% of the field genebank. The results of the present study provide essential information for the management and utilization of the germplasm collection. Single-tree genotyping of every tree in this collection is strongly recommended.