Modeling Agrobacterium-Mediated Gene Transformation of Tobacco (Nicotiana tabacum)-A Model Plant for Gene Transformation Studies

The multilayer perceptron (MLP) topology of an artificial neural network (ANN) was applied to create two predictor models in Agrobacterium-mediated gene transformation of tobacco. Agrobacterium-mediated transformation parameters, including Agrobacterium strain, Agrobacterium cell density, acetosyringone concentration, and inoculation duration, were assigned as inputs for ANN-MLP, and their effects on the percentage of putative and PCR-verified transgenic plants were investigated. The best ANN models for predicting the percentage of putative and PCR-verified transgenic plants were selected based on basic network quality statistics. Ex-post error calculations of the relative approximation error (RAE), the mean absolute error (MAE), the root mean square error (RMS), and the mean absolute percentage error (MAPE) demonstrated the prediction quality of the developed models when compared to stepwise multiple regression. Moreover, significant correlations between the ANN-predicted and the actual values of the percentage of putative transgenes (R-2 = 0.956) and the percentage of PCR-verified transgenic plants (R-2 = 0.671) indicate the superiority of the established ANN models over the classical stepwise multiple regression in predicting the percentage of putative (R-2 = 0.313) and PCR-verified (R-2= 0.213) transgenic plants. The best combination of the multiple inputs analyzed in this investigation, to achieve maximum actual and predicted transgenic plants, was at OD600 = 0.8 for the LB4404 strain of Agrobacterium x 300 mu mol/L acetosyringone x 20 min immersion time. According to the sensitivity analysis of ANN models, the Agrobacterium strain was the most important influential parameter in Agrobacterium-mediated transformation of tobacco. The prediction efficiency of the developed model was confirmed by the data series of Agrobacterium-mediated transformation of an important medicinal plant with low transformation efficiency. The results of this study are pivotal to model and predict the transformation of other important Agrobacterium-recalcitrant plant genotypes and to increase the transformation efficiency by identifying critical parameters. This approach can substantially reduce the time and cost required to optimize multi-factorial Agrobacterium-mediated transformation strategies.