Enhancing petunia tissue culture efficiency with machine learning: A pathway to improved callogenesis

Area of Science:

• Plant Biotechnology
• Computational Biology
• Agricultural Science

Background:

• Petunia callogenesis is unpredictable and genotype-dependent, hindering efficient regeneration.
• Machine learning (ML) offers a powerful approach to analyze callogenesis data and predict optimal conditions.

Purpose of the Study:

• Develop a predictive ML model for petunia callogenesis.
• Optimize phytohormone concentrations to enhance callus formation rate (CFR) and callus fresh weight (CFW).

Main Methods:

• Compared three ML algorithms: MLP, RBF, and GRNN.
• Utilized a genetic algorithm (GA) for phytohormone concentration optimization.
• Validated model predictions through laboratory experiments.

Main Results:

• GRNN demonstrated superior accuracy (R2≥83) compared to MLP and RBF.
• IBA was identified as the most influential phytohormone, followed by NAA, BAP, and KIN.
• Optimized phytohormone combination yielded a 95.83% CFR, validated experimentally.

Conclusions:

• The study presents a novel ML, sensitivity analysis, and GA approach for petunia callogenesis.
• Optimized phytohormone concentrations significantly improve callus formation.
• Findings support advancements in plant tissue culture and genetic engineering.