Machine learning is transforming all areas of biological science and industry, but it is typically restricted to a small number of users and scenarios. Tobias Erb and his colleagues at the Max Planck Institute for Terrestrial Microbiology created METIS, a modular software system for optimizing biological systems. The research team uses a variety of biological examples to demonstrate its applicability and versatility.
© Max-Planck-Institute for Terrestrial Micobiology/Bobkova |
Though biological system engineering is critical in biotechnology and synthetic biology, machine learning is now useful in all fields of biology. However, it is obvious that the application and improvement of algorithms, which are computational procedures composed of lists of instructions, is difficult. They are often limited not only by programming skills, but also by a lack of experimentally-labeled data. There is a need for efficient approaches to bridge the gap between machine learning algorithms and their applications for biological systems at the intersection of computational and experimental works.
Though biological system engineering is critical in biotechnology and synthetic biology, machine learning is now useful in all fields of biology. However, it is obvious that the application and improvement of algorithms, which are computational procedures composed of lists of instructions, is difficult. They are often limited not only by programming skills, but also by a lack of experimentally-labeled data. There is a need for efficient approaches to bridge the gap between machine learning algorithms and their applications for biological systems at the intersection of computational and experimental works.
Less information is required.
Active learning, also known as optimal experimental design, employs machine learning algorithms to suggest the next set of experiments after being trained on previous results, which is a useful approach for wet-lab scientists, particularly when working with a limited number of experimentally labeled data. However, one of the major bottlenecks is the experimentally labeled data generated in the lab, which is not always of sufficient quality to train machine learning models. "While active learning reduces the need for experimental data, we investigated various machine learning algorithms. We discovered a model that is even less dependent on data, which is encouraging "Amir Pandi, one of the study's lead authors, says
The team used METIS for a variety of applications to demonstrate its versatility, including protein production optimization, genetic constructs, combinatorial engineering of enzyme activity, and CETCH, a complex CO2 fixation metabolic cycle. They investigated a combinatorial space of 1025 conditions with only 1,000 experimental conditions for the CETCH cycle and reported the most efficient CO2 fixation cascade described to date.
Biological system optimization
The research provides novel tools for democratizing and advancing current efforts in biotechnology, synthetic biology, genetic circuit design, and metabolic engineering. "METIS enables researchers to optimize either previously discovered or synthesized biological systems," says Christoph Diehl, co-lead author of the study. "However, it also serves as a combinatorial guide for comprehending complex interactions and hypothesis-driven optimization. And, perhaps most importantly, it can be a very useful system for prototyping new-to-nature systems."
METIS is a modular tool that runs as Google Colab Python notebooks and can be accessed through a personal copy of the notebook in a web browser, without the need for installation, registration, or local computational power. The materials included in this work can help users.
Source: Materials provided by Max-Planck-Gesellschaft.