Design of tissue culture media for efficient Prunus rootstock micropropagation using artificial intelligence models

Establishing optimized protocols for micropropagation of some economical plants, such as Prunus sp., is still one of the most important challenges for in vitro plant culture researchers. As an example, micropropagation of GF677 hybrid rootstocks (peach × almond) are extremely dependent on the medium ingredients and a large undesirable proportion of GF677 shoots need to be discarded as a result of hyperhydricity and chlorosis. In this study, an artificial intelligence technique—specifically neurofuzzy logic—has been employed, as a modeling tool, to increase knowledge on the effect of 8 ion macronutrients (NH₄ ⁺, NO₃ ^?, Ca²⁺, K⁺, Mg²⁺, SO₄ ^2?, PO₄ ^2? and Na⁺; as inputs) on three growth parameters (outputs): total number of shoots per explant, healthy number of shoots per explant, and their bud number. The model delivered new insights, by three sets of IF–THEN rules, pinpointing the key role of NO₃ ^? and their interactions (NO₃ ^? × Ca²⁺ and NO₃ ^? × Ca²⁺ × K⁺) on all growth parameters measured. All growth parameters showed a high correlation ratio between experimental and predicted values being 77.48, 91.78 and 90.78 for total shoots, healthy number and bud number, respectively. Regression coefficients higher than 77 % together with statistical significant ANOVA (p < 0.01) indicated good performance of neurofuzzy logic models. Moreover, The model also can be used for inferring the best combination of ion concentrations to obtain high quality GF677 micropropagated shoots. In conclusion, we assess the utility of neurofuzzy logic technology in modeling complex databases, identifying new complex interactions among macronutrients, and inferring new results and valuable knowledge, which can be applied to design new plant tissue culture media and improve plant micropropagation.