Case Studies Using Machine Learning for Optimization of Cellular Factories To Produce Industrial Products

Edit This Case Study Record

	Using Machine Learning for Optimization of Cellular Factories To Produce Industrial Products

Using Machine Learning for Optimization of Cellular Factories To Produce Industrial Products

Technology Category	Analytics & Modeling - Machine Learning Analytics & Modeling - Predictive Analytics
Applicable Industries	Life Sciences Pharmaceuticals
Applicable Functions	Product Research & Development Quality Assurance
Use Cases	Predictive Maintenance Machine Condition Monitoring Remote Asset Management
Services	Software Design & Engineering Services System Integration Data Science Services
Challenge	Identify efficient tools to predict productivity in yeast cell factories and optimize protein pathways. Enabling accurate genotype-to-phenotype predictions through machine learning. Exploiting the power of combining mechanistic and machine learning models to effectively direct metabolic engineering efforts. Read More
About Customer	The Jensen Lab at DTU Biosustain, also known as Synthetic Biology Tools for Yeast (SBTY), develops engineering tools and high-resolution data that can dramatically increase the speed, efficiency, and rationale by which yeast-based cell factories are screened, selected, and engineered. Their particular focus is on how to engineer and evolve yeast genomes for improved performance of yeast as biocatalysts, and the development of high-throughput screening and selection tools enabled by genetically encoded biosensor devices. The Novo Nordisk Foundation Center for Biosustainability (DTU Biosustain) at the Technical University of Denmark is focused on developing new knowledge and technologies to help facilitate the transformation from the existing oil-based chemical industry to a more sustainable bio-based society in which chemicals are produced biologically. Read More
Solution	The partnership between DTU Biosustain and TeselaGen, in collaboration with the Berkeley Lab, aimed to identify the efficacy of using mechanistic models and machine learning models to enable accurate genotype-to-phenotype predictions. TeselaGen’s DISCOVER module provides an interface and compute infrastructure to create, train, and execute ML algorithms using TeselaGen’s Proprietary Software for fast data loading and processing. Once the genetic constructs of the study are designed, built, and experimentally characterized, various DISCOVER Machine Learning and/or Deep Learning tools created by TeselaGen can be applied. This approach allows for the deployment of powerful machine learning models, training predictive models, generating peptide leads, and running evolutive models, ultimately leading to optimized metabolic pathway designs and improved productivity in yeast cell factories. Read More Log in to view content
Contents

Technology Category

Analytics & Modeling - Machine Learning

Analytics & Modeling - Predictive Analytics

Applicable Industries

Life Sciences

Pharmaceuticals

Applicable Functions

Product Research & Development

Quality Assurance

Use Cases

Predictive Maintenance

Machine Condition Monitoring

Remote Asset Management

Services

Software Design & Engineering Services

System Integration

Data Science Services

Challenge

Identify efficient tools to predict productivity in yeast cell factories and optimize protein pathways. Enabling accurate genotype-to-phenotype predictions through machine learning. Exploiting the power of combining mechanistic and machine learning models to effectively direct metabolic engineering efforts.

About Customer

The Jensen Lab at DTU Biosustain, also known as Synthetic Biology Tools for Yeast (SBTY), develops engineering tools and high-resolution data that can dramatically increase the speed, efficiency, and rationale by which yeast-based cell factories are screened, selected, and engineered. Their particular focus is on how to engineer and evolve yeast genomes for improved performance of yeast as biocatalysts, and the development of high-throughput screening and selection tools enabled by genetically encoded biosensor devices. The Novo Nordisk Foundation Center for Biosustainability (DTU Biosustain) at the Technical University of Denmark is focused on developing new knowledge and technologies to help facilitate the transformation from the existing oil-based chemical industry to a more sustainable bio-based society in which chemicals are produced biologically.

Solution

The partnership between DTU Biosustain and TeselaGen, in collaboration with the Berkeley Lab, aimed to identify the efficacy of using mechanistic models and machine learning models to enable accurate genotype-to-phenotype predictions. TeselaGen’s DISCOVER module provides an interface and compute infrastructure to create, train, and execute ML algorithms using TeselaGen’s Proprietary Software for fast data loading and processing. Once the genetic constructs of the study are designed, built, and experimentally characterized, various DISCOVER Machine Learning and/or Deep Learning tools created by TeselaGen can be applied. This approach allows for the deployment of powerful machine learning models, training predictive models, generating peptide leads, and running evolutive models, ultimately leading to optimized metabolic pathway designs and improved productivity in yeast cell factories.

Impact #1	The Machine Learning approach recommended diverse candidate genes and cell factory designs with significant improvements compared to the original designs.
Impact #2	The collaboration enabled predictive strain engineering for high-performing results, achieving a GFP synthesis rate 106% higher than the improved platform design.
Impact #3	Optimal metabolic pathway designs were identified, resulting in improved titer and productivity of tryptophan by 74% and 43%, respectively, beyond the best cell factory designs used for training the algorithms.

Impact #1

The Machine Learning approach recommended diverse candidate genes and cell factory designs with significant improvements compared to the original designs.

Impact #2

The collaboration enabled predictive strain engineering for high-performing results, achieving a GFP synthesis rate 106% higher than the improved platform design.

Impact #3

Optimal metabolic pathway designs were identified, resulting in improved titer and productivity of tryptophan by 74% and 43%, respectively, beyond the best cell factory designs used for training the algorithms.

Benefit #1	GFP synthesis rate increased by 106%.
Benefit #2	Tryptophan titer improved by 74%.
Benefit #3	Tryptophan productivity increased by 43%.

Benefit #1

GFP synthesis rate increased by 106%.

Benefit #2

Tryptophan titer improved by 74%.

Benefit #3

Tryptophan productivity increased by 43%.

Download PDF Version

Overview

Using Machine Learning for Optimization of Cellular Factories To Produce Industrial Products

Operational Impact

Quantitative Benefit