Enhancing cellular production of fucoxanthin through machine learning assisted predictive approach in Isochrysis galbana

Abstract

The marine microalgae, Isochrysis galbana is a prolific producer of fucoxanthin which is a xanthophyll carotenoid with substantial global market value boasting extensive applications in the food, nutraceutical, pharmaceutical, and cosmetic industries. Although supplementation of different phytohormones to medium enhances fucoxanthin production, the quantification of pigment by conventional means is time-consuming and labor-intensive. This study addressed the multiple methodological limitations of HPLC-based fucoxanthin quantification and emphasized the need to develop a Machine Learning (ML) model as optimization and precise prediction remain a challenging task. Hence, an integrated experimental approach coupled with ML models was employed to predict fucoxanthin yield by supplementation of various phytohormones. The accuracy of fucoxanthin prediction excluding and including hormone descriptors was compared and evaluated using the ML models namely Random Forest (RF), Linear Regression (LR), Artificial Neural Network (ANN), and Support Vector Machine (SVM). RF model provided the most accurate prediction excluding hormone descriptors with coefficient of determination (R2 = 0.809) and root-mean-square error (RMSE = 0.776) followed by the ANN model with (R2 = 0.722) and (RMSE = 0.937). The inclusion of hormone descriptors for training and pre-processing of data further improved the fucoxanthin prediction accuracy of the RF model to (R2 = 0.839) and (RMSE = 0.712) and ANN model to (R2 = 0.738) and (RMSE = 0.909). These results indicated that the combination of low-cost, Ultraviolet (UV) spectrometric-based fucoxanthin quantification coupled with ML algorithms can be efficiently used for reliable prediction and enhanced fucoxanthin production, therefore highlighting a promising approach and furnishing invaluable insights towards the commercialization of microalgal fucoxanthin production.