Researchers from Chinese universities have developed an AI-powered platform that combines near-infrared (NIR) and Raman spectroscopy for real-time monitoring and control of antibiotic production, boosting efficiency by over 30%.
Spectroscopy Meets AI in the Fermentation Lab
In a major advancement for industrial biotechnology, a team of researchers from East China University of Science and Technology has integrated artificial intelligence (AI) with near-infrared (NIR) and Raman spectroscopy to monitor and optimize bioprocesses in real-time. Their innovative system dramatically improves the efficiency of gentamicin C1a fermentation—an antibiotic widely used in healthcare—demonstrating the promise of intelligent spectral sensing for industrial applications (1).
The study, published in Bioresource Technology, was led by Feng Xu, Lihuan Su, Hao Gao, Yuan Wang, Rong Ben, Kaihao Hu, Ali Mohsin, Chao Li, Ju Chu, and Xiwei Tian. The work was conducted across several affiliated institutions, including the State Key Laboratory of Bioreactor Engineering and the Qingdao Innovation Institute of East China University of Science and Technology, as well as the National Center of Bio-Engineering & Technology and the Shanghai Collaborative Innovation Center for Biomanufacturing Technology (1).
AI and dual-sensor spectroscopy supercharge antibiotic fermentation © fadi-chronicles-stock.adobe.com
Why Real-Time Monitoring Matters
In the realm of industrial fermentation, monitoring and regulating the environment within a bioreactor is crucial. Minor deviations in parameters such as glucose concentration or biomass growth can derail the entire process, leading to lower yields or compromised product quality. Traditional methods require manual sampling and lab analysis, which introduces delays and risks of contamination (1–3).
To address these limitations, the research team combined NIR and Raman spectroscopy—two complementary, non-invasive optical techniques—to track substrate and product concentrations in real-time. While NIR spectroscopy excels at monitoring bulk chemical features, Raman spectroscopy can detect specific molecular signatures, even at low concentrations (1–3).
Smarter Sensing Through Integration
Using machine learning (ML) algorithms, the researchers fused NIR and Raman datasets to overcome the shortcomings of single-sensor systems. Their hybrid model dramatically improved the predictive accuracy for key bioprocess parameters, with model performance increasing by up to 100.4% in terms of the coefficient of determination (R²). External validations showed R² values above 0.99, a clear indicator of the system’s robustness under varying fermentation conditions (1).
The team tested 13 different ML algorithms—including ridge regression, gradient boosting, and multilayer perceptron—to refine the prediction framework. Ultimately, the model was capable of generating accurate results within one minute, compared to over two hours required for conventional lab-based methods (1).
Autonomous Fermentation in Action
To put the system to the test, the researchers implemented it in a fed-batch fermentation process for producing gentamicin C1a using the Micromonospora echinospora 49-92S KL01 strain. Their AI-powered platform dynamically adjusted glucose feeding in response to real-time sensor data, keeping the glucose level stable at 5 g/L with less than 2% variation (1).
This precise control significantly enhanced the overall fermentation output. The final concentration of gentamicin C1a reached 346.5 mg/L—a 33% increase compared to traditional intermittent feeding strategies. Importantly, this improvement came with greater sustainability and reduced waste (1).
Implications for the Biotech Industry
The success of this dual-sensor, AI-driven approach demonstrates its potential for broader application in biomanufacturing. By enabling near-instantaneous feedback and control, this technology aligns well with the principles of Process Analytical Technology (PAT) and Quality by Design (QbD), which are increasingly favored in pharmaceutical production (1).
Furthermore, this method addresses a long-standing challenge in industrial biotechnology: how to maintain efficiency, scalability, and precision in the face of complex and variable biological systems. The study highlights the power of combining spectroscopy and AI, pushing industrial fermentation into a new era of precision and productivity (1–3).
References
(1) Xu, F.; Su, L.; Gao, H.; Wang, Y.; Ben, R.; Hu, K.; Mohsin, A.; Li, C.; Chu, J.; Tian, X. Harnessing Near-Infrared and Raman Spectral Sensing and Artificial Intelligence for Real-Time Monitoring and Precision Control of Bioprocess. Bioresour. Technol. 2025, 132204. DOI: 10.1016/j.biortech.2025.132204
(2) Rozov, S. Machine Learning and Deep Learning Methods for Predictive Modelling from Raman Spectra in Bioprocessing. arXiv 2020, arXiv:2005.02935. DOI: 10.48550/arXiv.2005.02935
(3) Esmonde-White, K. A.; Cuellar, M.; Lewis, I. R. The Role of Raman Spectroscopy in Biopharmaceuticals from Development to Manufacturing. Anal. Bioanal. Chem. 2022, 414, 969–991. DOI: 10.1007/s00216-021-03727-4
Toward a Generalizable Model of Diffuse Reflectance in Particulate Systems
June 30th 2025This tutorial examines the modeling of diffuse reflectance (DR) in complex particulate samples, such as powders and granular solids. Traditional theoretical frameworks like empirical absorbance, Kubelka-Munk, radiative transfer theory (RTT), and the Hapke model are presented in standard and matrix notation where applicable. Their advantages and limitations are highlighted, particularly for heterogeneous particle size distributions and real-world variations in the optical properties of particulate samples. Hybrid and emerging computational strategies, including Monte Carlo methods, full-wave numerical solvers, and machine learning (ML) models, are evaluated for their potential to produce more generalizable prediction models.
