This research investigates advanced computational modeling strategies for predicting Limnospira platensis growth dynamics under mixotrophic cultivation with the use of agro-industrial processing residues. The cultivation approach combines photosynthetic carbon fixation with organic carbon utilization, potentially enhancing both biomass yields and valuable metabolite production. By implementing sophisticated data analysis frameworks, this work aims to optimize cultivation parameters while advancing waste-to-resource conversion principles. Three effluents, cheese whey (scotta), tomato extract, and artichoke extract, were tested at multiple concentrations and compared with a control medium. Experimental results revealed that whey-based substrates achieved the highest biomass productivity, reaching up to 6.1 g L−1, while artichoke extract media resulted in lower yields (approximately 3.0 g L−1). To optimize predictive performance, Gradient Boosting algorithms were tuned through Bayesian hyperparameter optimization, demonstrating strong predictive capabilities for biomass accumulation, with an overall determination coefficient of R2 = 0.97 and a root mean square error of 0.19 in cross-validation. Feature importance analysis identified temporal parameters and conversion efficiency as predominant factors governing biomass accumulation. These findings demonstrate that the integration of data-driven modeling framework and circular bioeconomy strategies can effectively transform agro-industrial waste streams into valuable cost-efficient substrates, reducing waste and supporting sustainable production of high-value biomolecules for food, pharmaceutical, and nutraceutical applications.
Advanced feature engineering and gradient boosting optimization for predicting Limnospira platensis growth dynamics under mixotrophic conditions using agro-industrial byproducts
Bartolomeo Cosenza
;
2025-01-01
Abstract
This research investigates advanced computational modeling strategies for predicting Limnospira platensis growth dynamics under mixotrophic cultivation with the use of agro-industrial processing residues. The cultivation approach combines photosynthetic carbon fixation with organic carbon utilization, potentially enhancing both biomass yields and valuable metabolite production. By implementing sophisticated data analysis frameworks, this work aims to optimize cultivation parameters while advancing waste-to-resource conversion principles. Three effluents, cheese whey (scotta), tomato extract, and artichoke extract, were tested at multiple concentrations and compared with a control medium. Experimental results revealed that whey-based substrates achieved the highest biomass productivity, reaching up to 6.1 g L−1, while artichoke extract media resulted in lower yields (approximately 3.0 g L−1). To optimize predictive performance, Gradient Boosting algorithms were tuned through Bayesian hyperparameter optimization, demonstrating strong predictive capabilities for biomass accumulation, with an overall determination coefficient of R2 = 0.97 and a root mean square error of 0.19 in cross-validation. Feature importance analysis identified temporal parameters and conversion efficiency as predominant factors governing biomass accumulation. These findings demonstrate that the integration of data-driven modeling framework and circular bioeconomy strategies can effectively transform agro-industrial waste streams into valuable cost-efficient substrates, reducing waste and supporting sustainable production of high-value biomolecules for food, pharmaceutical, and nutraceutical applications.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
