Predictive modeling of antioxidant activity in Syzygium malaccense leaf extracts using image processing and machine...

Area of Science:

• Phytochemistry
• Computational Chemistry
• Pharmacognosy

Background:

• *S. malaccense*, a plant from the Myrtaceae family, is traditionally used and recognized for its high flavonoid and phenolic compound content.
• Understanding the antioxidant potential of medicinal plants is crucial for developing new therapeutic agents.

Purpose of the Study:

• To evaluate the antioxidant potential of *S. malaccense* leaf extracts and fractions.
• To apply machine learning models for predicting antioxidant properties based on assay images.

Main Methods:

• Antioxidant assays including DPPH, ABTS, and Ferric Reducing Antioxidant Power (FRAP).
• Spectroscopic analysis and image processing of assay results.
• Machine learning models (SVM, decision tree, Random Forest, XGBOOST, LightGBM, CatBoost) applied to image data.
• Generalized Linear Model (GLM) analysis for solvent effects.

Main Results:

• XGBOOST and Random Forest models showed high performance (R² up to 99.35% in training, 95.50% in testing).
• GLM analysis indicated acetate solvent yielded the highest FRAP, while hexane yielded the lowest.
• Ethanol extraction was optimal for ABTS radical scavenging, and dichloromethane for DPPH radical scavenging.

Conclusions:

• Machine learning models, particularly XGBOOST and Random Forest, show significant promise for quantifying plant antioxidant properties using image analysis.
• The study validates *S. malaccense* as a source of antioxidants and demonstrates an innovative method for their assessment.