Embedding response surfaces in kinetic models predicts microalgae growth across light and nutrient gradients
Abstract
Microalgae are promising cell factories for sustainable production of biofuel, high-value compounds, and wastewater bioremediation. However, their industrial deployment requires robust prediction of biomass productivity across the wide range of light and nutrient conditions encountered in photobioreactors. Mechanistic models typically assume constant growth parameters, limiting their accuracy across varying cultivation conditions. Conversely, response surface methodology (RSM) captures parameter variability across conditions but lacks dynamics and can yield biologically implausible predictions under untested cultivation conditions. This study introduces a hybrid model that embeds biologically interpretable Monod-type surfaces into a logistic growth model, allowing maximum growth rate and carrying capacity to vary independently and continuously with light intensity and nutrient availability. Validated on 1315 growth curves of Chlamydomonas reinhardtii grown mixotrophically, the model predicts full time series under previously untested light–nutrient combinations at 2 scales: flask (50 mL, R² = 0.94) and microplate (200 µL, R² = 0.84), consistently outperforming three established mechanistic models. Six global parameters replace up to 50 condition-specific ones, substantially reducing the experimental effort. We further demonstrate that microplate-derived parameter surfaces can be mapped onto flask-scale surfaces through simple transformations (R² ≥ 0.94), providing a basis for scale-up from high-throughput data to bench-scale cultivation. The framework offers a practical tool to optimize microalgal cultivation at laboratory scale, with preliminary evidence supporting its transferability across culture volumes.
Related
- Do protein representations encode variant function? Diagnosing residue-level protein attributions via embedding-space gradients
- Large-Scale Network Embedding Predicts Opposing Functions of Adjacent Regions Within Human Prefrontal Cortex
- Response of Non-Structural Carbohydrates in Picea crassifolia to Elevation Gradients in the Qilian Mountains
- Prefrontal activation predicts response latency and is shaped by age and lifestyle
- The fraction of CD117/c-KIT-expressing erythroid precursors predicts ESA response in low-risk myelodysplastic syndromes.
BibTeX
@article{pietri2026embedding,
title = {{Embedding response surfaces in kinetic models predicts microalgae growth across light and nutrient gradients}},
author = {Mélanie Pietri and Thomas Rodet and Bruno Le Pioufle and Thomas Nowak and Matthias Függer and Sakina Bensalem},
journal = {Bioresource Technology},
year = {2026},
month = oct,
doi = {10.1016/j.biortech.2026.135364},
url = {https://hal.science/hal-05774563v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Embedding response surfaces in kinetic models predicts microalgae growth across light and nutrient gradients
AU - Mélanie Pietri
AU - Thomas Rodet
AU - Bruno Le Pioufle
AU - Thomas Nowak
AU - Matthias Függer
AU - Sakina Bensalem
PY - 2026
DA - 2026/10/02
JO - Bioresource Technology
DO - 10.1016/j.biortech.2026.135364
UR - https://hal.science/hal-05774563v1
AB - Microalgae are promising cell factories for sustainable production of biofuel, high-value compounds, and wastewater bioremediation. However, their industrial deployment requires robust prediction of biomass productivity across the wide range of light and nutrient conditions encountered in photobioreactors. Mechanistic models typically assume constant growth parameters, limiting their accuracy across varying cultivation conditions. Conversely, response surface methodology (RSM) captures parameter variability across conditions but lacks dynamics and can yield biologically implausible predictions under untested cultivation conditions. This study introduces a hybrid model that embeds biologically interpretable Monod-type surfaces into a logistic growth model, allowing maximum growth rate and carrying capacity to vary independently and continuously with light intensity and nutrient availability. Validated on 1315 growth curves of Chlamydomonas reinhardtii grown mixotrophically, the model predicts full time series under previously untested light–nutrient combinations at 2 scales: flask (50 mL, R² = 0.94) and microplate (200 µL, R² = 0.84), consistently outperforming three established mechanistic models. Six global parameters replace up to 50 condition-specific ones, substantially reducing the experimental effort. We further demonstrate that microplate-derived parameter surfaces can be mapped onto flask-scale surfaces through simple transformations (R² ≥ 0.94), providing a basis for scale-up from high-throughput data to bench-scale cultivation. The framework offers a practical tool to optimize microalgal cultivation at laboratory scale, with preliminary evidence supporting its transferability across culture volumes.
SN - 0960-8524
ER -