Neural networks models as decision-making tool for in vitro proliferation of hardy kiwi

Plant micropropagation is a complex procedure involving the interaction of several factors such as mineral media composition, vitamins, plant growth regulator, etc. Indeed, almost each species or cultivar seems to need specific culture conditions to achieve the highest proliferation rate. The aim of this work was to establish which of the next four factors: culture media, number of subcultures, and concentration of two plant growth regulators (BAP and GA(3)), cause the most significant effect on Actinidia arguta (Sieb. & Zucc.) Planch. ex Miq. multiplication. In this paper, neural network models were built and used as decision-making tools. Arguta kiwi apical explants were cultured in three commonly kiwifruit (Actinidia sp.) micropropagation media [Murashige and Skoog (MS), Standardi (St) and Gamborg's (B5)] containing different concentrations of BAP (0, 0.5, 1.0, 1.5, 2.0 and 2.5 mg L-1) and GA(3) (0, 0.1, 0.5 and 1.0 mg L-1) during three consecutive subcultures. Two response parameters were measured in this work: shoot number and shoot length. Database was analyzed using the neurofuzzy logic technique that allowed a simple, predictive and easy to interpret model to be obtained. The model disclosed that the key factors were BAP concentration for shoot number and the combination of BAP and GA(3) for shoot length. In fact, if A. arguta explants were cultivated with intermediate or high BAP (> 0.6 mg L-1) concentration, the highest number of shoots was obtained. But, low concentration of BAP (0-0.3 mg L-1) and high of GA(3) (0.5-1 mg L-1) should be needed for long shoots. The model also indicated that the number of subcultures is another key factor, being necessary at least two subcultures to observe improvement on the shoot number, but only one to achieve the longest shoots. Finally, the model pointed out that all media caused a positive effect on shoot length but only St a slight positive in shoot number, suggesting that their composition need to be optimized. Our results point out the use of neurofuzzy logic as decision-making tool and its utility on obtaining knowledge of the effects of the factors on the plant micropropagation.