Nature Communications: “Combining mechanistic and machine learning models for predictive engineering and optimization of tryptophan metabolism"

Algorithms can engineer cells for you

Saturday 26 Sep 20

Contact

Michael Krogh Jensen
Senior Researcher
DTU Biosustain
+45 61 28 48 50

Berkeley Lab scientists in collaboration with DTU Biosustain researchers develop and train a tool that could drastically speed up the ability to design new biological systems.

In a new study published in Nature Communications, Jie Zhang and Søren Petersen of the Novo Nordisk Foundation Center for Biosustainability at DTU in collaboration with a team of international scientists used and trained algorithms to predict productivity in yeast cell factories.

 

This training resulted in recommended cell factory designs with impressive improvements compared to the original cell factory designs, explains Senior Researcher Michael Krogh Jensen at The Novo Nordisk Foundation Center for Biosustainability (DTU Biosustain):

 

“This approach enabled us to successfully forward-engineer the aromatic amino acid metabolism in yeast, with the new recommended designs ultimately improving titer and productivity by 74% and 43%, respectively, beyond the best cell factory designs used for training the algorithms. The merger of machine learning algorithms and advanced synthetic biology tools is expected to move engineering of living cells from a painstaking trial-and-error approach to a more cost-effective approach leveraging predictive simulations of cell metabolism," he said.

 

ART: The Automated Recommendation Tool

In the Nature Communications study: “Combining mechanistic and machine learning models for predictive engineering and optimization of tryptophan metabolism”, the team at DTU Biosustain used an algorithm, ART (Automated Recommendation Tool) developed by Lawrence Berkeley National Laboratory (Berkeley Lab), to guide the metabolic engineering process to increase the production of tryptophan, an amino acid with various uses, by baker’s yeast Saccharomyces cerevisiae.

 

The project was led by Jie Zhang and Søren Petersen in collaboration with scientists at Berkeley Lab and Teselagen, a San Francisco-based startup company. To conduct the experiment, they selected five genes based on an advanced genome-scale model developed at Chalmers Tekniska Högskola.

 

By specifying 6 different expression levels for each of the five genes a library of metabolic pathway designs spanning nearly 8,000 potential combinations was constructed. The DTU Biosustain researchers then used sequencing, physiology and biosensors to obtain experimental data on 250 of those pathway designs, representing just 3% of all possible combinations. This data was used to train a variety of machine learning algorithms, which ultimately enabled ART to recommend pathway designs with desired outputs (amino acid production) based on presented input (gene expression).

 

Then, using statistical inference, the tool was able to extrapolate how all possible remaining combinations would affect tryptophan production. The best design ultimately recommended increased tryptophan production (expression) by 74% and 43%, respectively, beyond the best cell factory designs used for training the algorithms.

 

"The merger of machine learning algorithms and advanced synthetic biology tools is expected to move engineering of living cells from a painstaking trial-and-error approach to a more cost-effective approach leveraging predictive simulations of cell metabolism"
Senior Researcher Michael Krogh Jensen

From 150 person-years to weeks?

Scientists at Berkeley Lab have developed ART, that adapts machine learning algorithms to the needs of synthetic biology to guide development systematically.

 

The innovation means scientists will not have to spend years developing a meticulous understanding of each part of a cell and what it does in order to manipulate it; instead, with a limited set of training data, the algorithms are able to predict how changes in a cell’s DNA or biochemistry will affect its behavior, then make recommendations for the next engineering cycle along with probabilistic predictions for attaining the desired goal.

 

“The possibilities are revolutionary,” said Hector Garcia Martin, a researcher in Berkeley Lab’s Biological Systems and Engineering (BSE) Division. “Right now, bioengineering is a very slow process. It took 150 person-years to create the anti-malarial drug, artemisinin. If you’re able to create new cells to specification in a couple weeks or months instead of years, you could really revolutionize what you can do with bioengineering.”

 

In “ART: A machine learning Automated Recommendation Tool for synthetic biology,” led by Berkeley Lab data scientist Tijana Radivojevic, the researchers presented the algorithm, which is tailored to the particularities of the synthetic biology field: small training data sets, the need to quantify uncertainty, and recursive cycles. The tool’s capabilities were demonstrated with simulated and historical data from previous metabolic engineering projects, such as improving the production of renewable biofuels.

 

Synthetic biology has the potential to make significant impacts in almost every sector: food, medicine, agriculture, climate, energy, and materials. The global synthetic biology market is currently estimated at around $4 billion and has been forecast to grow to more than $20 billion by 2025, according to various market reports.

 

Part of this article is with courtesy to the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) and was kindly lent to us from a press release published on 25 September 2020. Please go to the original press release to learn more about ART and machine learning in bioengineering.