Synthetic biology is a rapidly growing field, allowing scientists to design biological systems to specification, such as for artificial meat, biosensors or biological drug delivery platforms. The global synthetic biology market is currently estimated at around $4bn and has been forecast to grow to more than $20bn by 2025.

However, conventional bioengineering methods are laborious, requiring scientists to spend years developing a detailed understanding of each part of a cell and its function, involving lengthy trial and error processes.

Now, researchers from Berkeley Lab have developed a tool (the Automated Recommendation Tool, aka ART) which uses a machine learning algorithm to systematically guide the development of new biological systems. The algorithm is adapted for the needs of synthetic biology, with small training data sets, the need to quantify uncertainty and recursive cycles.

Using a limited set of training data, this tool could predict how changes in a cell’s DNA or biochemistry will impact its behaviour and then make recommendations for its next engineering cycle.

“The possibilities are revolutionary,” said Hector Garcia Martin, lead author of the Nature Communications study. “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 [of] weeks or months instead of years, you could really revolutionise what you can do with bioengineering.”

The new tool was demonstrated with simulated and historical data from previous metabolic engineering projects. Garcia Martin and his colleagues used ART to guide the metabolic engineering process to increase production of an amino acid, tryptophan, by a species of yeast.

They selected five genes, each controlled by various mechanisms and totalling nearly 8,000 potential combinations of biological pathways. They then collected experimental data on 250 of the pathways (three per cent of all possible combinations) and used this data to train an algorithm to identify what amino acid production is associated with which gene expression.

Using statistical inference, the tool was able to extrapolate how many of the remaining combinations would affect the production of tryptophan. The design recommended by ART more than doubled its production over the reference strain.

Garcia Martin commented: “This is a clear demonstration that bioengineering led by machine learning is feasible and disruptive if scalable. We did it for five genes, but we believe it could be done for the full genome.

“This is just the beginning. With this, we’ve shown that there’s an alternative way of doing metabolic engineering. Algorithms can automatically perform the routine parts of research while you devote your time to the more creative parts of the scientific endeavour: deciding on the important questions, designing the experiments and consolidating the obtained knowledge.”

Synthetic biology is a rapidly growing field, allowing scientists to design biological systems to specification, such as for artificial meat, biosensors or biological drug delivery platforms. The global synthetic biology market is currently estimated at around $4bn and has been forecast to grow to more than $20bn by 2025.

However, conventional bioengineering methods are laborious, requiring scientists to spend years developing a detailed understanding of each part of a cell and its function, involving lengthy trial and error processes.

Now, researchers from Berkeley Lab have developed a tool (the Automated Recommendation Tool, aka ART) which uses a machine learning algorithm to systematically guide the development of new biological systems. The algorithm is adapted for the needs of synthetic biology, with small training data sets, the need to quantify uncertainty and recursive cycles.

Using a limited set of training data, this tool could predict how changes in a cell’s DNA or biochemistry will impact its behaviour and then make recommendations for its next engineering cycle.

“The possibilities are revolutionary,” said Hector Garcia Martin, lead author of the Nature Communications study. “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 [of] weeks or months instead of years, you could really revolutionise what you can do with bioengineering.”

The new tool was demonstrated with simulated and historical data from previous metabolic engineering projects. Garcia Martin and his colleagues used ART to guide the metabolic engineering process to increase production of an amino acid, tryptophan, by a species of yeast.

They selected five genes, each controlled by various mechanisms and totalling nearly 8,000 potential combinations of biological pathways. They then collected experimental data on 250 of the pathways (three per cent of all possible combinations) and used this data to train an algorithm to identify what amino acid production is associated with which gene expression.

Using statistical inference, the tool was able to extrapolate how many of the remaining combinations would affect the production of tryptophan. The design recommended by ART more than doubled its production over the reference strain.

Garcia Martin commented: “This is a clear demonstration that bioengineering led by machine learning is feasible and disruptive if scalable. We did it for five genes, but we believe it could be done for the full genome.

“This is just the beginning. With this, we’ve shown that there’s an alternative way of doing metabolic engineering. Algorithms can automatically perform the routine parts of research while you devote your time to the more creative parts of the scientific endeavour: deciding on the important questions, designing the experiments and consolidating the obtained knowledge.”