Integrating Machine Learning and Microwave-Assisted Green Extraction: Total Colorimetric Response Assay-Based Optimization of Opuntia ficus-indica Seed Residues

What if the secret to high-value antioxidants was hiding in the seeds you throw away? A new study uses machine learning and microwaves to unlock them in record time.

Prickly pear seeds—often discarded as waste—are quietly packed with phenolic and flavonoid compounds. Researchers decided to give these leftovers a second life, using microwave-assisted extraction (MAE) to pull out the good stuff efficiently and sustainably.

But this wasn't a simple shake-and-bake. The team designed a multi-step optimization strategy, starting with single-factor experiments, then response surface methodology (Box–Behnken design), and finally a machine learning model called K-nearest neighbors coupled with the dragonfly algorithm (KNN_DA). They tested how ethanol concentration, microwave power, time, and liquid-to-solid ratio affected the extracts' colorimetric responses and antioxidant activity.

The winning formula: 50% ethanol, 800 W microwave power, 4 minutes, and a liquid-to-solid ratio of 47.28 mL/g. Under these conditions, the Folin–Ciocalteu reducing capacity hit 376.85 mg GAE/100 g dry weight, and the AlCl3 complexation response reached 49.16 mg QE/100 g DW, with prediction errors under 3%. The optimized extracts showed enhanced antioxidant activity, proving MAE is fast, green, and effective—and that KNN_DA is a powerful ally for process optimization.

Key Points

• Microwave-assisted extraction rapidly recovers antioxidants from prickly pear seed waste.
• Machine learning model KNN_DA predicts optimal extraction conditions with <3% error.
• Best conditions: 50% ethanol, 800 W, 4 min, 47.28 mL/g liquid-to-solid ratio.