Nutrient based classification of Phyllospora comosa biomasses using machine learning algorithms: Towards sustainable...

Area of Science:

• Marine Botany
• Biochemistry
• Data Science

Background:

• Sustainable seaweed value chains require precise biochemical characterization for product development and quality assurance.
• Australian seaweed species are underutilized, lacking comprehensive biochemical investigations for industrial applications.

Purpose of the Study:

• To characterize the biochemical composition of Phyllospora comosa thallus segments (blades, stipes, vesicles) and whole samples from Victoria, Australia.
• To develop and apply machine learning models for accurate classification of seaweed collection sites and segments.

Main Methods:

• High-throughput characterization techniques were employed to analyze nutrient profiles (carbohydrates, minerals, amino acids).
• Machine learning classification models (rpart) were utilized to differentiate collection sites based on cadmium levels and segments based on glutamic acid and potassium content.

Main Results:

• Phyllospora comosa biomass is rich in carbohydrates (64-68%), ash (27-31%), potassium, sodium, calcium, magnesium, and iodine.
• Significant variations were observed: stipes are carbohydrate-rich, blades are high in glutamic acid, calcium, magnesium, and iodine, and vesicles are rich in potassium.
• Machine learning achieved 88% accuracy in separating collection sites by cadmium levels and 100% accuracy in classifying segments by glutamic acid and potassium content.

Conclusions:

• Accurate biochemical characterization and machine learning classification of Phyllospora comosa are crucial for its valorization.
• Segment-specific nutrient profiles suggest diverse applications, enabling targeted product development.
• These methods can enhance the authenticity and sustainability of Australian seaweed value chains, promoting diversification and expansion.