Machine learning-driven modelling and optimization of callus induction and biomass accumulation in Lavandula ×...

Area of Science:

• Plant Biotechnology
• Agricultural Science
• Computational Biology

Background:

• Callus formation in Lavandula × intermedia is influenced by explant type, plant growth regulators (PGRs), and culture duration.
• The combined effects of these factors on callus induction and biomass accumulation are not fully understood.
• Optimizing callus production is crucial for efficient propagation and secondary metabolite extraction in Lavandula × intermedia.

Purpose of the Study:

• To comprehensively evaluate the effects of 57 PGR treatments on callus formation from root and stem explants of Lavandula × intermedia.
• To model and predict callus induction and biomass accumulation using statistical and machine learning approaches.
• To identify optimal conditions for maximizing callus biomass and induction efficiency.

Main Methods:

• In vitro culture of Lavandula × intermedia explants with 57 different PGR treatments over 15 weeks.
• Application of five statistical and machine learning models (XGBoost, Random Forest, etc.) for data analysis and prediction.
• Feature importance analysis and multi-objective optimization using NSGA-II to identify key factors and optimal conditions.

Main Results:

• Callus induction was rapid, typically within three weeks, and biomass accumulation followed a biphasic pattern, peaking after six weeks.
• A combination of 0.5 mg/L 2,4-D and 0.5 mg/L kinetin yielded the highest biomass (up to 35 g).
• XGBoost demonstrated high predictive accuracy (R² ≈ 0.94), while Random Forest offered stable performance; culture duration was the dominant factor for biomass.

Conclusions:

• Machine learning integration enables robust prediction and optimization of callus formation in Lavandula × intermedia.
• Culture duration significantly impacts biomass accumulation, with hormonal composition playing a key role.
• The study provides valuable insights for optimizing micropropagation protocols and enhancing secondary metabolite production in lavender.