Pleistocene dispersion supports a unique native diversity in the Colombian avocado germplasm

Genomic characterization of ex-situ collections optimizes the utilization of genetic resources, identifies redundancies among accessions, captures cryptic variation, establishes reference collections, and ultimately assists pre-breeding and breeding efforts. However, the integration of population genomic analyses is often lacking when studying the biodiversity of crop gene pools. Here, we present modern classification and machine learning approaches to characterize and harness the genebank of an agrobiodiversity hotspot on Persea americana Mill., an iconic American fruit tree crop that has seen an unprecedented expansion worldwide. We selected 144 accessions from the Colombian National genebank and 240 materials from local plantations in the Colombian Northwest Andes. We genotyped them using a strategy based on reduced representation sequencing. We included available sequences of genotypes from known avocado races, Mexican, Guatemalan, and West Indian, to discover SNPs, analyze the population structure and identify possible new genetic groups in Colombian germplasm. We detected a population structure suggestive of a new fourth race in Colombia, with a possible genetic substructure related to geographical origin (Andean and Caribbean). Hybrid determination and ABC modeling suggested rampant inter-race geneflow. They supported the hypothesis of the high mobility of native avocado trees from Central America to Northern South America starting in the Pleistocene. Our study supports that Colombia might be a new diversity center for P. americana. Genotypes of the two newly identified Colombian groups can be used as parents in plant breeding strategies to generate cultivars adapted to specific ecogeographical regions of Colombia.