Predictive modeling-guided sequential fermentation enhances phenolic biotransformation and antioxidant potential of wet Spirulina platensis beverage

High susceptibility of wet Spirulina platensis to post-harvest degradation causes significant nutritional and functional quality losses. Fermentation is a proven strategy to improve bioactive properties but the advancement of this process through integration of predictive modeling analysis remains uninvestigated. Therefore, this study combined response surface methodology (RSM) with Multilayer perceptron artificial neural network (MLP-ANN) modeling to improve the accumulation of phenolic compounds in wet Spirulina beverage via sequential fermentation with Kluyveromyces marxianus and Lactiplantibacillus plantarum . Predicted fermentation conditions (yeast-to-bacteria inoculum ratio of 3.75:1.25, glucose concentration of 8.5%, and Spirulina concentration of 9.5%) suggested the production of 60.56 mg gallic acid equivalents (GAE)/L of total phenolic content (TPC)，which was experimentally confirmed at 60.86 mg GAE/L. Both RSM and MLP-ANN models demonstrated strong predictive performance (R 2  > 0.99), but RSM outperformed MLP-ANN by achieving 0.49% of relative error as compared to 0.74% for MLP-ANN. Experimental validation of predictive condition confirmed 76.2%, 119.6%, and 96.69% increase in TPC, total flavonoid content (TFC), and DPPH (1,1-diphenyl-2-picrylhydrazyl) radical scavenging activity respectively. UPLC-MS (ultra-performance liquid chromatography-mass spectrometry) analysis of TPC revealed substantial increases in caffeic acid, kaempferol, and apigenin contents in the fermented beverage. TPC improvement could potentially be attributed to the possible activation of complementary microbial biotransformation pathways including aromatic amino acid metabolism and flavonoid precursor synthesis. An improvement in sensory acceptability and reduction in detectable biogenic amines has also been observed under predicted sequential fermentation conditions. These findings underscore the feasibility of predictive models to improve the functional quality of Spirulina -based fermented beverages.