Abstract

Over-Engineering in the Life Sciences: Microfluidics and Microphysiological Systems Meet Artificial Intelligence

John Wikswo, A.B. Learned Professor of Living State Physics; Founding Director, Vanderbilt Institute for Integrative Biosystems, Vanderbilt University

It is a worthwhile exercise to examine whether the addition of microfluidic pumps and valves to organ-chip systems represents either “over-engineering” or the means by which microphysiological systems can be made massively parallel. While several cited examples of over-engineering represent overly complicated or frivolous solutions to a problem, such as the $400 Juicero juice-pouch-squeezer or a $1,390 Porsche roof box rated at 200 km/hour, others are prescient and ultimately yield great value, for example the NASA Opportunity rover whose mission was planned to last 90 Mars sols (~93 Earth days) but in fact lasted 14 Earth years, and Mercedes Benz’s introduction in the 1970s of antilock braking systems and in 1980 the driver’s airbag and seat-belt tensioner. From the perspective of a biologist who uses an agar-filled Petri dish for microbial colony picking, a 1536 well plate and its associated high-throughput screening infrastructure are overkill; for big Pharma, it is a welcome tool that took 15 years to be refined and put into practical use. As we approach the 15th anniversary of Mike Shuler’s landmark 2007 patent “Devices and Methods for Pharmacokinetic-based Cell Culture System,” we realize that microfluidic tissue and organ chips have been perfused using gravity, pressurized reservoirs, external syringe or peristaltic pumps, or on-chip peristaltic pumps. Although the smallest devices abound with Quake-style pneumatic valves that also serve as pumps, very few organ-chip systems utilize valves. For over a decade, our group has been developing microfluidic pumps and valves, and we have long argued that multi-organ microphysiological systems will ultimately need automation of pumps and valves to control each chip, their closed-loop sensors, and organ-organ interconnections. Our Missing Organ MicroFormulator concept spawned an award-winning 96-channel MicroFormulator, which served as the foundation for CN Bio Innovations’ recently introduced PhysioMimixTM pharmacokinetic (PK) system that supports PK studies of oncological drugs. As our group develops its fourth generation of rotary microfluidic pumps and valves to enable our construction and Ross King’s coding of the machine-learning algorithms for the 1000-channel Genesis Self-Driving Chemostat for yeast systems biology, we realize that this hardware can be readily adapted to perfuse and control a dozen of Steve George’s and Scott Simon’s bone marrow and skin chips in an in vitro infection model, maintain and analyze multiple, coupled organ chips for six months or longer, support remote studies of select agents in a BSL-3 or BSL-4 facility, apply circadian rhythms to thousands of wells in dozens of well plates or hundreds of zebrafish or perfused organ chips, and accelerate and optimize the microbial production of antibodies, industrial feed stocks, food protein, and sequestered carbon dioxide. Is this over-engineering or the future of artificial intelligence in biology?