Bioproduction Culture Quality Control Using Day-Specific ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biological production of cell cultures in bioreactors is slow, expensive, and subject to failure and variability due to data variability and incompleteness, making it challenging to build accurate predictive models.
Innovation Solution
Implement machine learning models trained on process-specific historical data for each culture day, using dimension reduction and ensemble techniques like Gaussian Mixture Models and recurrent neural networks to provide real-time quality control recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional predictive models are used with historical data from multiple bioreactors and runs, then production outcomes can be predicted, but data variability and incompleteness significantly reduce model accuracy
Solution Approach 1:
The patent segments the historical data by creating separate training datasets for each bioreactor unit. Each bioreactor's data is processed independently to build unit-specific predictive models, avoiding the noise and variability introduced by aggregating data from multiple different bioreactors. This segmentation approach maintains data consistency and improves prediction accuracy for each specific unit.
2Reliability
If production runs are extended to ensure successful outcomes, then product quality is maintained, but production time and costs increase
Solution Approach 1:
The patent implements preliminary action by using predictive models to forecast production outcomes before the actual production run completes. The system continuously monitors real-time data and compares it against predicted trajectories, enabling early detection of potential failures. This allows producers to take corrective actions or terminate unsuccessful runs early, avoiding unnecessary extension of production cycles and reducing overall production time while maintaining success rates.
3Reliability
If comprehensive monitoring of all process parameters is implemented, then production failures can be detected, but system complexity and measurement costs increase
Solution Approach 1:
The patent extracts and focuses on monitoring only the most critical process parameters that have the greatest impact on production outcomes. Rather than comprehensively monitoring all possible parameters, the system identifies key indicators through analysis of historical data and predictive modeling, then concentrates monitoring resources on these essential parameters. This extraction approach maintains reliable failure detection capability while significantly reducing system complexity and measurement costs.
Data Source
AI summary
Real-time quality control of a culture for bioproduction is facilitated using machine learning. In this approach, real-time process data for a set of parameters for a current production run is received. Based on this process data, a prediction is made using an instance of a machine learning model that has been trained on process data from past production or development runs. The instance is uniquely associated to a particular culture day and thus independent of any other instance of the machine learning model (for other culture days). Based on the prediction, a quality control recommendation for the current production run is then made. Several different types of predictions are enabled, and various different recommendations are provided based on the predictions.