Use of multiple regression analysis and artificial neural networks to model the effect of nitrogen in the organogenesis of Pinus taeda L.

Abstract

Mineral nutrition is a very important factor in the success of in vitro plant cultures. The aim was to compare the predictive capacity of the models obtained using a parametric technique such as multiple regression analysis with a nonparametric one such as artificial neural networks. These techniques were used for modeling the effect of total nitrogen concentration and the ratio nitrate: ammonium in the regeneration rate, oxidation rate, callus proliferation rate, number of buds per explant and buds-forming capacity index. Both the concentration of total nitrogen and the relationship between the concentrations of nitrate and ammonium influenced the morphogenetic responses. Optimal buds regeneration was in the range of 10–20 mM of the total nitrogen concentration and 1–2 of the nitrate: ammonium ratio. Higher concentrations of nitrogen produced an increase in the oxidation rate while the low nitrate: ammonium ratio favored the callus proliferation rate. Artificial neural network models presented a better precision to predict the different responses to the total content of nitrogen and the nitrate: ammonium rate, with higher coefficients of determination and correlation. They also presented a lower root mean square error for all the variables studied than the multiple regression analysis. The use of artificial neural networks allows obtaining a better model of the effect of nitrogen on the organogenesis of Pinus taeda L. than traditional statistical techniques.