Artificial neural network and decision tree facilitated prediction and validation of cytokinin-auxin induced in vitro organogenesis of sorghum (Sorghum bicolor L.)

In this study, in vitro regeneration protocol of sorghum ( Sorghum bicolor ) was successfully established by using direct organogenesis from a mature zygotic embryo explant. The used basal medium encompassed Murashige and Skoog medium (MS) supplemented with 2–4 mg/L Benzylaminopurine (BAP) alone or with 0.25 mg/L Indole butyric acid (IBA) or Naphthalene acetic acid (NAA). Results demonstrated a significant impact of cytokinin-auxin on shoot count (1.24–3.46) and shoot length (2.80–3.47 cm). Maximum shoot count (3.46) and shoot length (3.97 cm) were achieved on the MS medium enriched with 2 mg/L BAP + 0.25 mg/L NAA and 2.0 mg/L BAP, respectively. To ascertain the impact of BAP alone, BAP + IBA, and BAP + NAA, the data were also analyzed by using a factorial regression model. Pareto chart and normal plots were used to check either the positive or negative impact of input variables on output variables. To further explore the association between BAP + IBA and BAP + NAA on shoot count and shoot length, contour and surface plots were also built. Three different artificial intelligence-based models along with four different performance metrics were utilized to validate the predicted results. Multilayer perceptron (MLP) model performed more efficiently ( R 2 = 0.799 for shoot count and R 2 = 0.831 for shoot length) as compared to the decision tree-based algorithms of random forest (RF) – ( R 2 = 0.779 for shoot count and R 2 = 0.786 for shoot length) and extreme gradient boost (XGBoost) – ( R 2 = 0.768 for shoot count and R 2 = 0.781 for shoot length). As plant tissue culture protocol is a powerful tool for genetic engineering and genome editing of crops, integration of different artificial intelligence-based models can lead to improvement of sorghum with the aid of biotechnological tools. Graphical abstract Key message Optimization of In vitro whole plant regeneration of sorghum using mature zygotic embryo explant and cytokinin-auxin combination Data analysis through ANOVA, factorial regression anlaysis and machine learning Presentation of data via Pareto charts, standarized plots, contour plots and surface plots Better performance of ANN-based MLP model compared to decision tree based RF and XGBoost model.