Sustainable beetroot bioactives extraction with NADES-MAE: A spider algorithm-optimized hybrid SVR-XGBoost model

Beetroot is packed with antioxidants, but getting them out sustainably has always been a puzzle. Now, scientists have turned to a spider-inspired algorithm to crack it.

Beetroot, a humble root vegetable, is a treasure trove of phenolics and flavonoids—compounds with potent antioxidant power. Yet, extracting these bioactives efficiently without harming the environment is no small feat. In a new study, researchers paired natural deep eutectic solvents (NADES) with microwave-assisted extraction (MAE), creating a green extraction method that's both effective and eco-friendly.

To fine-tune the process, the team ran 30 systematic experiments, varying microwave power, temperature, time, and solvent ratio. They measured total phenolic content, total flavonoid content, and antioxidant activity. Then came the clever part: a hybrid machine learning model combining support vector regression (SVR) and extreme gradient boosting (XGBoost) was trained to predict yields and antioxidant activity.

Two optimization strategies were tested: standard cross-validation and a spider optimization algorithm. The spider-inspired approach consistently outperformed, achieving test R² values of 0.75 for TPC, 0.77 for TFC, and 0.68 for DPPH%. The model also revealed that different bioactives require different predictive weights—SVR dominated TPC prediction, while TFC needed a balanced ensemble.

This research demonstrates that combining eco-friendly solvents with intelligent algorithms can lead to sustainable, efficient extraction processes. The validated models identified distinct optimal conditions for each response, highlighting the need for output-specific optimization. This platform holds strong potential for industrial applications, paving the way for greener production of bioactive compounds.

Key Points

• Green extraction of beetroot bioactives using NADES-MAE
• Hybrid SVR-XGBoost optimized by spider algorithm outperforms cross-validation
• Output-specific optimal conditions identified for TPC, TFC, DPPH%