Pollen performance modelling with an artificial neural network on commercial stone fruit cultivars

Abstract

Pollen tube growth and pollen germination percentage are key factors for successful fruit set. Pollen performance is critical for the production and breeding of flowering plants and in agricultural systems in terms of fruit development. This study was carried out to predict pollen tube growth and pollen germination percentage in four stone fruits species (cherry (Prunus avium), apricot (Prunus armeniaca), plum (Prunus domestica), and peach (Prunus persica)) using a neural network. For this purpose, we measured pollen tube length and pollen germination rates under in vitro conditions. For the in vitro test, pollen grains of four stone fruit cultivars were sown in three different media and incubated at seven different temperatures for four incubation periods. A layered neural network was used for estimating the pollen germination rate and pollen tube length related to the in vitro condition. This study suggests a method for estimating the pollen germination rate and pollen tube length using artificial neural networks. The performed artificial neural networks produced an efficient prediction from in vitro data. The determination coefficients obtained between the observed and predicted data sets are 0.86 (for germination rate) and 0.81 (for tube length), indicating an accurate estimation of the in vitro data. In our case, the network that produced the best result had a 4:9:2 architecture.