Modeling and optimizing callus growth and development in Cannabis sativa using random forest and support vector...

Area of Science:

• Plant Biotechnology
• Computational Biology
• Cannabis sativa Research

Background:

• Plant callus, undifferentiated cells, is vital for metabolite production and regeneration.
• Callogenesis is complex, influenced by factors like plant growth regulators (PGRs).
• Machine learning offers potential for modeling and optimizing non-linear biological processes like callus development.

Purpose of the Study:

• To evaluate PGR effects on Cannabis sativa callus morphology.
• To maximize callus growth and promote embryogenic callus production.
• To apply machine learning for predicting and optimizing callogenesis.

Main Methods:

• Image processing combined with random forest (RF) and support vector machine (SVM) algorithms.
• SVM was selected for higher predictive accuracy over RF.
• SVM was integrated with a genetic algorithm (GA) for PGR optimization.

Main Results:

• SVM demonstrated superior accuracy in modeling callus traits compared to RF.
• The SVM-GA model accurately predicted optimal PGR levels for callus growth.
• A significant correlation was found between embryogenic callus production and true callus density.

Conclusions:

• Machine learning, particularly SVM-GA, effectively models and optimizes PGRs for Cannabis sativa callus development.
• Optimized PGR levels can enhance callus growth and embryogenic production.
• Callus density is a key indicator for predicting embryogenic potential.