No one would sell us a high-throughput system that could be reconfigured in minutes and would scale to fill a factory. So we built one.
In the past year, Zymergen has nearly doubled the size of our high-throughput life sciences facility, and our need for lab automation has grown even faster. Zymergen’s labs run experiments that last from weeks to months, use a remarkable diversity of organisms, and target dozens of end products. We onboard new protocols and processes every week.
In the past, we managed our scale by building integrated automation platforms, where a central robot arm was used to transport samples between devices like thermal cyclers and centrifuges. These platforms were powerful, but they couldn’t keep up with our requirements: they were expensive, difficult to prototype, and time-consuming to reconfigure. Given the ever-evolving nature of our workflows, fast prototyping and reconfigurability were crucial.
We needed something new — laboratory automation that would be easy to maintain, fit into our space-constrained labs, and adapt to our rapidly-changing operations. The Zymergen Automation group has created it: a modular system of Reconfigurable Automation Carts (RACs).
Each RAC contains a scientific instrument, like a liquid handler or centrifuge, along with its own dedicated robot arm and a segment of magnetic track. These segments of track can be attached together to join multiple RACs, and plates can travel along the connected track. This provides a simple and modular way to string together a series of devices required to run a complex biological workflow.
The modular connections between RACs allow us to rearrange devices in the span of minutes, rather than needing to take down an integrated system for hours or days. We can easily adapt our equipment to changing processes and adjust its capacity across a wide range of program scales. We can automate earlier and experiment more freely. It’s changing how we think about automating science.
Finding the Right Level of Automation
A laboratory isn’t simply automated or not. There are many levels of automation:
- Benchtop: includes equipment like a thermal cycler or plate sealer sitting on a laboratory bench that can make protocols faster or more repeatable.
- Walk-away: involves more complex devices, such as liquid handlers, that can run through a programmed set of instructions while a scientist works elsewhere in the lab.
- Integrated: requires devices connected by a robot arm that can automate more complex protocols. This can allow, for example, a microplate to be centrifuged and weighed before it is measured by a plate reader.
Walk-away and integrated systems free up our scientists to spend more time on interesting questions and less time on the repetitive mixing of translucent liquids. They increase our throughput in other ways, too, leading to speedier dispensing and more repeatable results.
So why isn’t our entire lab composed of islands of integrated automation?
First, it takes a long time to build out integrated automated systems. Formalizing protocols, running rigorous equivalence testing, strategizing about how to fit large robots into small spaces, and figuring out how to connect new equipment in established (and often crowded) labs add layers of challenges.
Second, an integrated system, once built, is hard to change. To fully automate our protocols, we need to connect a diverse set of devices, but a central arm’s limited reach forces us to pack them too closely for easy maintenance. Even if we put the arm on a rail, it eventually becomes a bottleneck and prevents us from using each device at capacity. In addition, teaching and aligning robot arms can be a huge pain; for all but a few arms, a five-millimeter nudge during a repair can result in plates of very expensive microbes being spilled on very expensive instruments.
This inflexibility (and the associated expenses) means that companies often don’t adopt integrated automation until they’ve developed locked-down, mature workflows. What’s more, once a lab starts to automate, the cost and difficulty of experimentation forces workflows to stay static. Not only is the automation constrained by the maturity of one’s workflows, but workflows become constrained by the automation.
Due to these challenges, we believe that most labs, including ours, are under-automated. Zymergen has processes at many levels of maturity — it would have been a huge effort for us to invest solely in traditional automation methods when our protocols change every few months. At the same time, we knew that all of our projects could benefit from the standardization, precision, reproducibility, and speed of automation.
From Inflexible to Modular
As we tried to bring our labs to the right level of automation, we found that the limitations of centralized robot arms were often the greatest obstacle to reconfiguring our integrated systems. We ultimately realized that we needed to fully decentralize our plate transport — rather than relying on one or two fragile, expensive central arms that were time-consuming to align, we chose to pair modular sections of magnetic track with low-cost, dedicated robot arms. This breakthrough led to the design of the RACs: each contains its own section of track and its dedicated robot arm only needs to reach a few locations within the cart. The teach points move with the cart; we can rearrange devices without having to re-teach a central robot arm.
The RACs also make maintenance much easier. Each RAC is on wheels, so we can roll a RAC out of the lab when we need to fix or tweak it — our service engineers no longer have to compete for space with busy scientists. Because each section of the magnetic track is individually actuated, wheeling away a cart won’t necessarily bring down the whole system — we can mind the gap by routing our plates around it or by filling it with a spare section of track.
Most of the time, we’re not just going to wheel away a RAC — we want to swap in a new one. We gave a lot of thought to designing standard connectors between the carts that would be able to line up the track without too much precise fiddling. With two engineers and two power drills, our current design enables us to disconnect one cart and swap in a replacement in just ten minutes.
This added flexibility and robustness doesn’t just help with maintenance. When we’re experimenting with a new protocol or onboarding a new microbe, we can connect two or three carts, run a sub-process on their devices, and iterate on both process and automation without being locked into a design. When a protocol becomes out of date, we can disassemble the system that ran it and quickly hand off the carts to someone else.
When we need to make do with unconventional lab dimensions (thank you, Bay Area real estate) we’re not tied to any particular geometry — we can even run track through a tunnel in a wall if necessary. When we need to add capacity, we can simply add another RAC or two onto the end of the track system. When we need to add redundancy, we can add them in series or run workflows in parallel.
The RAC design provides a standard for plate transport and connections; it doesn’t make any assumptions about which devices are on the cart. This means we can integrate both standard laboratory tools and bespoke ones. What’s more, because we contain our geometric complexity within each RAC, we can add new capabilities and capacity without a linear increase in hardware complexity. Our system is one of the first that facilitates reconfigurability at large scales.
The reconfigurability of this system means avoiding stranded CapEx, and easing the adoption of new workflows in a scientific field that’s both highly varied and rapidly evolving. There’s no pressure, financial or otherwise, to automate a process all at once. It means we can always be at the right level of automation.
Learning What We Can Build, Now That We Have Tools
We’ve designed a new method for modular, configurable, scalable automation. Now we’re excited to learn what we can do with it.
One of our senior hardware engineers likes to talk about LEGO® (I mean, who among us doesn’t?). He points out that you can’t start thinking about how to piece together an overambitious LEGO pirate galley or rocket ship until you’ve figured out the basics of how the bricks connect.
Earlier this year, a connected system of RACs in one of our protocol development labs ran Zymergen’s first 24-hour batch of PCR runs, transporting plates to and from storage as they finished their thermal cycling. In a few weeks, we’ll be releasing a system of RACs to one of our most high-throughput strain optimization labs. This system integrates ten different kinds of devices and will be able to run through many of our most common workflows while our scientists design experiments, analyze data, and (because it can run 24 hours a day) get a good night’s rest.
We’re learning how the RAC bricks connect with our science, and how we can use them to build more complexity without sacrificing flexibility. These Reconfigurable Automation Carts are already helping our scientists and engineers routinely try out new ideas in automation and workflows.
Traditional laboratory automation couldn’t keep up with Zymergen’s operations. Modular automation can; what’s more, this technology is expanding what we think is possible in our lab. We’re thinking about what we can now piece together, and we’re excited to eventually show you what we build.
Do you have an interest in integrated, modular automation? We’d love to hear feedback on (or questions about) what we’ve done so far. You can reach us at email@example.com.
Tessa Alexanian is an Engineer on the Automation Platform Software team at Zymergen.