If you’ve eaten meat-flavored vegan burgers or used cosmetics with synthetic collagen, then synthetic biology has benefited you. Both of these products are “grown” in the laboratory, and this is also an area of development with great potential. It allows scientists to create biological systems with specific specifications, such as creating a microbe to produce an agent to fight cancer. However, traditional bioengineering methods are slow and labor-intensive, with trial and error being the main approach. To solve the problem, scientists at the US Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) have developed a new tool that adapts machine learning algorithms to the needs of synthetic biology. This will systematize the development management. The researchers reported their findings in the journal Nature Communications.
The new development will allow scientists not to spend years on a detailed understanding of each part of the cell and its specific functions in order to control it. Instead, with a limited set of training data, new algorithms will predict how changes in DNA will affect cell behavior and biochemistry, and then provide recommendations for the next engineering cycle, along with plausible predictions to achieve the desired goal for engineers.
“The possibilities are revolutionary,” said Hector Garcia Martin, a researcher in the Biological Systems and Engineering (BSE) division of the Berkeley lab who led the study. “Bioengineering is currently a very slow process. It took 150 man-years to develop the antimalarial drug, artemisinin. If you can create new cells to specifications in a couple of weeks or months instead of a few years, that will revolutionize the possibilities of bioengineering. ”
Working with BSE data analyst Tijana Radivojevic and an international team of researchers, the team developed and demonstrated a patent-pending algorithm – the Automatic Recommendation Tool (ART). Machine learning allows computers to make predictions after being “trained” based on a significant amount of available data.
The algorithm is adapted to the peculiarities of the field of synthetic biology with its small training data sets, the need to quantify uncertainty, and recursive loops. The capabilities of this tool have been demonstrated using simulations and historical data from previous metabolic engineering projects such as improving the production of renewable biofuels.