Inteligência artificial aplicada ao estudo da diversidade genética do Banco de Germoplasma de videira da EMBRAPA Semiárido

Abstract

The objective of this work was to apply artificial intelligence, through the use of the ESOM clustering algorithm, to the study of the genetic diversity of genotypes of table grapes from the Embrapa Semiárido Vine Germplasm Active Bank. The following quantitative agronomic variables of 93 table grape genotypes were evaluated: yield (kg/plant), number of clusters per plant, cluster weight (g), cluster length and width (cm), berry weight (g), length and diameter of berries (mm), soluble solids content (ºBrix), titratable acidity (g/100ml) and soluble solids/titratable acidity ratio. The experiment was carried out in the Experimental Field of Mandacaru, Juazeiro-BA, and data referring to four harvesting seasons (2018.1, 2018.2, 2019.1 and 2019.2) were analyzed. The grouping performed by the ESOM network was able to discover genetic patterns and differences between the studied table grape genotypes, allowing the formation of 10 heterotic groups. Group 0 had the highest maximum values for the variables yield (8.44 kg), cluster weight (520.19 g), cluster length (22.30 cm), cluster width (15.10 cm), weight of berry (8.00 g), berry length (26.84 mm), berry diameter (22.17 mm) and soluble solids (23.90 °Brix) and also presented the highest mean values for cluster weight (331.20 g), cluster length (16.6 cm), cluster width (10.5 cm), berry weight (5.78 g), berry length (23.5 mm) and berry diameter (19.7 mm). The presence of genetic variability for the analyzed variables was evidenced by the formation of genetic variability maps, demonstrating a broad genetic base for the 93 genotypes analyzed. The variables with less variability were cluster length and berry length. The ESOM matrix of genetic similarity indicated the most promising crosses among the 10 heterotic table grape groups based on genetic divergence. The crossing of genotypes from group 0, mostly composed of genotypes with seeds and large berries, with genotypes from group 1, 2, 4, 5 and 7, which are seedless and have smaller berries being the most indicated. Crosses between genotypes from groups 0 and 9 are not indicated. Despite being the most genetically distant groups, all individuals in group 9 have a small berry size and have seeds, characteristics considered undesirable in a table grape breeding program. ESOM proved to be promising in the analysis of genetic diversity and consequent formation of heterotic groups, in addition to indicating which crosses are more promising. Future studies on the validation of ESOM as an efficient grouping method for plant genetic improvement are indicated.