Modeling and studying in vitro regeneration in common bean breeding using artificial neural networks and machine learning algorithms

In the realm of biotechnological enhancement of common beans, an imperative challenge lies in devising a reliable and effective in vitro regeneration strategy, given the inherent difficulty of regenerating this crop in laboratory settings. This research, aiming to address this challenge, leverages the power of Machine Learning (ML) models, specifically employing algorithms for Artificial Neural Networks (ANN). The primary objective is to establish an efficient and repeatable in vitro regeneration process while simultaneously optimizing and predicting future outcomes. The study incorporates various variables such as bean genotype, explants, and different doses of 6-benzylaminopurine (BAP) and CuSO4. A Recurrent Regression Neural Network (RRNN) is employed to model and anticipate the results of in vitro crop regeneration, specifically focusing on common beans. The experimental setup involves preconditioning common bean embryos with 10, 15, and 20 mg/L BAP for 25 days, followed by growth in a post-treatment environment comprising 0.3, 0.6, 0.9, and 1.2 mg/L BAP for 7 weeks. Subsequently, the plumular apice is isolated for in vitro regeneration. Notably, the RRNN model is also integrated with a Genetic Algorithm (GA) to optimize the regeneration process further. The results are compelling, with RRNN exhibiting the lowest Mean Squared Error (MSE) of 0.061, signifying superior predictive accuracy in total regeneration. In comparison, Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGB) models exhibit higher MSE values of 0.081, 0.081, and 0.097, respectively. These findings underscore the efficacy of the RRNN algorithm, outperforming other models across all parameters. The superior performance of RRNN suggests its potential application in making precise predictions regarding common bean regeneration. In the context of a common bean breeding program, these outcomes can be harnessed to optimize and predict plant tissue culture methods, thereby enhancing biotechnological techniques employed in the cultivation of common beans. The integration of ML models, particularly RRNN, stands as a promising avenue for advancing crop regeneration strategies and contributing to the efficiency of biotechnological interventions in agriculture.