Cell Culture Scale-Up Prediction Using Learned Process Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cell culture processes require a long period due to the need for multiple experiments to optimize parameters in transitioning from small-scale initial culture to large-scale production culture, limiting the efficiency of cell culture operations.
Innovation Solution
A program and apparatus utilizing machine learning to derive a learned model that predicts culture results in large-scale production processes based on initial culture data, enabling efficient transition and optimization of culture conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If intermediate culture processes with multiple experiments are performed to optimize parameters, then culture conditions can be optimized for production, but the time required from initial culture to production culture increases significantly
Solution Approach 1:
The system performs preliminary analysis by training a machine learning model on historical culture data from both initial and production culture processes. This pre-trained model can then predict optimal production culture conditions based on initial culture results, eliminating the need for time-consuming intermediate experiments and directly providing optimized parameters for production scale-up.
2Loss of time
If machine learning is used to predict production culture results from initial culture data, then the transition period can be shortened, but the complexity of the system increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between initial culture data and production culture predictions. This model acts as a mediator that learns the complex relationships from historical data and provides predictions without requiring direct complex experimental procedures, thus reducing time while managing complexity through data-driven abstraction.
Solution Approach 2:
The system creates a virtual copy of the production culture process through machine learning modeling. By training the model on historical production culture data, it generates a digital twin that can predict outcomes without requiring physical intermediate experiments, thereby shortening the transition period while the complexity is contained within the computational model rather than physical experimental infrastructure.
3Reliability
If traditional experimental methods are used to optimize culture parameters, then reliable optimization can be achieved, but the number of experiments required increases the time and resource consumption
Solution Approach 1:
The system implements feedback by continuously training the machine learning model on historical culture data from both initial and production processes. The model learns from past experiment results and uses this feedback to improve its predictions, maintaining reliability while reducing the need for new experiments. The feedback loop allows the system to accumulate knowledge over time, improving accuracy without increasing current experimentation burden.
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A program for operating a cell culture support apparatus causes a computer to function as a first acquisition unit, a second acquisition unit, and a first derivation unit. The first acquisition unit acquires a learned model indicating a relationship between an initial culture process and a production culture process, derived by performing machine learning on the basis of initial record data for learning including a set of culture condition data indicating a record of culture conditions and culture result data indicating a record of culture results in the initial culture process, and production record data for learning, corresponding to the initial record data for learning and including a set of culture condition data indicating a record of culture conditions and culture result data indicating a record of culture results in the production culture process. The second acquisition unit acquires initial record data for analysis including a set of the culture condition data indicating the record of the culture conditions and the culture result data indicating the record of the culture results in the initial culture process, and temporary culture condition data indicating temporary culture conditions in the production culture process. The first derivation unit derives predicted culture result data obtained by predicting the culture result data in the production culture process, from the learned model acquired in the first acquisition unit, and the initial record data for analysis and the temporary culture condition data acquired in the second acquisition unit.