Predicting Metabolic Pathway Dynamics with Machine Learning, w/ Zak Costello - #163

Metabolic Engineering

Zak Costello provides a fascinating analogy to explain metabolic engineering, likening it to a subway system where proteins act as tunnels connecting various chemical stations. This process involves adding reactions to an organism to produce new chemicals, such as biofuels, by borrowing proteins from diverse life forms like yeast and trees 1. Zak's journey from electrical engineering to synthetic biology highlights the interdisciplinary nature of this field. He emphasizes the role of machine learning in handling the complexity of synthetic biology, allowing him to contribute effectively despite his non-biological background 2.

The central dogma of biology is kind of important here, which is DNA is transcribed into RNA, RNA is translated into proteins. In metabolic engineering, proteins are used to catalyze reactions.

--- Zak Costello

This integration of machine learning with synthetic biology aims to make biological systems predictably engineerable, addressing challenges like creating biofuels and biomaterials.

   

Biohacking

Biohacking has become surprisingly accessible, with tools like CRISPR allowing hobbyists to experiment with genetic modifications. Zak Costello explains that biohacker spaces enable enthusiasts to work with microorganisms and plants, making genetic modifications as simple as transferring liquids 3. In microbial engineering, proteins from various organisms are integrated into E. coli to produce compounds like limonene. This process involves extracting genetic codes from organisms like mint and yeast and inserting them into E. coli, which acts as a macro machine for producing desired compounds 4.

CRISPR, just like all the things we were talking about today, it's a protein, and so it's defined by a piece of genetic code.

--- Zak Costello

This accessibility opens up new possibilities for innovation and experimentation in bioengineering.

   

ML in Biology

Machine learning plays a crucial role in synthetic biology by predicting metabolic pathway dynamics. Zak Costello describes using time series data to model the dynamic behavior of metabolic systems, akin to reconstructing a hill from the trajectories of a rolling ball 5. This approach allows for the prediction of system characteristics using traditional machine learning models like random forests, which help in understanding the relationship between metabolic states and their derivatives 6.

We figure with time series data, what you have is this relationship for how something moves.

--- Zak Costello

By integrating these models over time, researchers can simulate cellular behaviors, enhancing the efficiency of strain development and reducing the need for extensive physical experimentation.