Machine learning - based optimization of integrated extraction method for R-phycoerythrin from dry biomass of Gracilaria corticata

Abstract

Macroalgal biomass, a valuable source of bioactive compounds, is highly perishable in its wet form. Drying helps in the extension of its shelf life. However, conventional wisdom indicates that drying increases the mass transfer resistance, making extraction of biomolecules from dry biomass more difficult. The present study is aimed at developing the most suitable method for the primary extraction of R- Phycoerythrin (R-PE) from dry biomass of Gracilaria corticata. Practically, such reports are not available. Different extraction methods alone and their integration (in order to enhance the yield) are employed. A pre-soaking (12h) step of the dry biomass prior to extraction is found to significantly decrease the mass transfer resistance in all the primary extraction methods. Key extraction process parameters such as solid-liquid ratio (1:20, 1:30, and 1:40), extraction time (10, 20, and 30 min), and ultrasonication amplitude (20, 40, and 60 %) were standardized. The Box-Behnken Design (BBD) is employed to identify the configurations covering the entire range of the selected parameters of the integrated extraction method. This approach enabled optimization through a reduced number of experiments. Further, a machine learning model, the Gradient boosting model, is employed to predict the yield of R-PE for all possible combinations of extraction process parameters. Among all the methods employed, ‘ultrasound-assisted extraction + homogenization’ resulted in the highest yield of R-PE (1.23 mg/g dw) with an extraction efficiency of 87.8 %.