Genetic patterns recognition in crop species using self-organizing map: the example of the highly heterozygous autotetraploid potato (Solanum tuberosum L.)

摘要

We tested the ability of the self-organizing map (SOM), a type of artificial neural network, in revealing genetic patterns within the autotetraploid potato (Solanum tuberosum L.). A total of 591 potato varieties, originating from various main European breeder collections and released into different market segments between 1815 and 2015, were examined using a set of 21 informative microsatellite markers. The consistency of this artificial intelligence approach in detecting genetic stratifications in such a homogeneous population was evaluated through the comparison with three other multivariate methods that are widely used for this purpose. Results showed that the SOM was equally suitable for classifying varieties into main detected groups and visualizing inter-group genetic dissimilarities. When it came to reveal the organization of the population structure at the intra-group level, traditional multivariate methods lost in resolution. Contrariwise, the SOM provided additional information on the intra-group diversity by highlighting a multitude of consistent subgroups, which seemed to be mainly related to their common heritage, spatio-temporal features and certain agronomic traits. Relations between computed SOM subgroups and the market segments were subject to certain elucidations. The relevance of using more flexible multivariate statistical approaches for mapping population structures of crop species is considered throughout this paper in terms of current and future prospects for breeding programs.