Project
Machine Learning Inspired Synthetic Biology: Neuromorphic Computing in Mammalian Cells

Advisors: Prof. Ron Weiss, BE, EECS  and Prof. Skylar Tibbits

Synthetic biologists seek to collect, refine, and repackage nature so that it’s easier to design new and reliable biological systems, typically at the cellular or multicellular level. While research in experimental synthetic biology has been fruitful with standardized devices, many inspired by the fields of electronics and integrated circuit design, one current challenge facing synthetic biology is the development of genetic networks which are both compact and specially designed to perform a unique task or fit a complex, custom behavior.

For instance, while cell classification circuits, which selectively treat cells according to their molecular makeup, have huge potential to impact personalized medicine, they are typically implemented using complex networks of binary logic gates, and must be re-engineered on a per-application basis. In contrast, in-silico gene expression classifiers are commonly built using strategies from artificial intelligence, which take better advantage of the information encoded in the analog levels of the bio-molecules of interest and require fewer computations to do so. In this thesis, I propose that synthetic analog gene circuits can be engineered to perform the same types of computation inside living systems. As proof of concept, I implement an analog perceptron in Human Embryonic Kidney cells which is able to develop a classification function of external cues through generational learning, and which can serve as a composable unit in more complex networks. I then demonstrate extensions of the circuit in silico which are capable of adapting their behavior without outside intervention.

While biology has long served as inspiration for the artificial intelligence (AI) community, this thesis contributes to a new, interactive relationship between the two fields. Here, more than just a helpful metaphor, nature offers to AI an active probe, helping us discover how learning systems work in living things, and beyond.