Clone-Specific Digital Twin Training for Bioprocess Kinetics

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Digital Twins are rarely applied in biotechnological production processes due to limited applicability across different organisms and cell lines, high measurement data requirements, and the need for new data and models for each application case, leading to significant resource and time costs, as well as reduced predictive quality.

Innovation Solution

A generalized Digital Twin is generated by training machine-learning models with experimental data across clones, cell lines, and process formats, combining generic and clone-specific components to improve predictive quality, allowing for reuse of data and model parameters across different application cases, and integrating these models with metabolic functionalities and reactor models for adaptable process setups.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If Digital Twins are applied in biotechnological production processes, then predictive quality and process control improve, but data requirements and measurement precision demands increase significantly

Engineering Contradiction:
Improvepredictive qualityVSAvoidmeasurement data requirements
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent creates a universal Digital Twin framework that works across different organisms, cell lines, and process formats through a standardized model structure. The generic model components can be adapted to multiple biotechnological applications without requiring complete redesign for each case, reducing the overall data burden while maintaining predictive quality.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The Digital Twin model is segmented into generic components that can be trained on aggregated data from multiple sources and clone-specific components that capture individual variations. This segmentation allows the system to leverage data efficiency from the generic parts while maintaining accuracy for specific applications.

Inventive Principle:
Principle #1Segmentation

2Reliability

If new data and models are generated for each application case, then model accuracy improves, but resource costs and time consumption increase significantly

Engineering Contradiction:
Improvemodel accuracyVSAvoidtime costs
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent performs preliminary training of generic model components using aggregated data from multiple application cases before deploying to specific applications. This pre-training establishes a solid foundation that reduces the amount of additional training needed for each new case, significantly reducing time costs while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent merges data from multiple application cases to train the generic model components, combining information across different organisms, cell lines, and processes. This merging creates a more robust baseline model that requires less application-specific data and training time.

Inventive Principle:
Principle #5Merging (Combining)

3Measurement precision

If clone-specific models are created for each organism, then prediction accuracy for that clone improves, but data requirements and computational costs increase

Engineering Contradiction:
Improveprediction accuracyVSAvoiddata requirements
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The model is segmented into generic components trained on aggregated data from multiple clones and process formats, and clone-specific components that capture individual variations. This segmentation allows the system to leverage data efficiency from the generic parts while maintaining accuracy for specific applications.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The generic model components serve multiple clones and application cases simultaneously, reducing the total data requirements compared to creating entirely separate models for each clone. The universal framework maintains predictive accuracy while being more data-efficient.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4296350A1A concept for training and using at least one machine-learning model for modelling kinetic aspects of a biological organism
Publication Date: 2023.12.27 YOKOGAWA INSILICO BIOTECHNOLOGY GMBH
  • EP4296350A1 patent drawingFigure 1
  • EP4296350A1 patent drawingFigure 2a~2c
  • EP4296350A1 patent drawingFigure 3~9

AI summary

Examples relate to a concept for training and using at least one machine-learning model for modelling kinetic aspects of a biological organism, and in particular to a method, apparatus, and computer program for training the at least one machine-learning model for modelling the kinetic aspects of the biological organism, and various methods using such a trained at least one machine-learning model. The method for training the at least one machine-learning model for modelling the kinetic aspects of the biological organism comprises training the machine-learning model based on training data. The training data is based on experimental data of a plurality of clones of the biological organism. The training data comprises a subset of training data that is based on experimental data of a single clone. A first component of the at least one machine-learning model is trained using the training data, with the first component representing a generic kinetic behavior of the biological organism. A second component of the at least one machine-learning model is trained using the subset of the training data, the second component representing a clone-specific kinetic behavior of the biological organism.