AI-generated from publicly available materials.Integrating low-fidelity data with high-fidelity experimental information may enhance the accuracy of machine learning models in biopharmaceutical process development, as suggested by recent research.
Machine learning holds promise for accelerating and optimizing biopharmaceutical processes by enabling predictions about the effects of parameter changes. However, the high-fidelity data necessary for training these models is costly and often scarce. Researchers propose that incorporating low-fidelity data—reflecting general trends—into training sets can address this limitation. Mohammad Golzarijalal, a research fellow at the University of Melbourne, emphasizes that while high-fidelity data fine-tunes model predictions, low-fidelity data can provide broader insights, helping models avoid overfitting and enhancing their predictive capacity across diverse conditions.
Improved models could lead to better forecasts of critical factors such as cell growth and product yield, ultimately aiding in optimizing various operational parameters. For instance, Golzarijalal cites a study where combining simulated data with experimental results led to more accurate predictions of monoclonal-antibody titers compared to models relying solely on experimental data.
While drug companies currently utilize low-fidelity data to some degree, its systematic application remains underexplored. Golzarijalal notes that multi-fidelity algorithms can facilitate the integration of historical data into current projects, enhancing the utility of past experiences for future predictions and decision-making. This approach could significantly transform how biopharmaceutical processes are modeled and optimized.