Advancing Biology with an Open-Source Robot
A little startup’s big ambitions to improve lab automation and revolutionize life sciences research
Tech ambition can be measured by what analogies the founders like to drop in interviews. The Uber of X, Amazon for Y, all pretty impressive, if the startup in question can pull it off. But Will Canine, the organizing force behind OpenTrons, is aiming even higher: “Biology lab automation today is where computers were in the 1960s — a big mainframe with punch cards, run by lab techs. We’re the PC.”
For anyone with an excitable futurist mindset, this calls up all kinds of insane techno-utopian fantasies. When you look at the Kickstarter for Opentrons’ open-source robot, the OT.One, you think: Finally, a MakerBot for life itself. Donate enough to the campaign (hoping that they actually deliver the thing), and you’ll be printing out glow-in-the-dark fish that smell like strawberries and excrete lifesaving cancer drugs in no time.
But what this actually means — a PC revolution for biology — is a little more complicated, and a lot more practical. The OT.One is, technically speaking, a liquid-handling robot. It doesn’t extrude oozing synthetic life out onto a plate, it automates the tedious work that makes up most of biology research. Canine’s mainframe analogy is spot-on, since existing bio lab robots cost, at their most basic, around $50,000 (rising quickly into the mid-six figures), require weeks of training to use properly, are decidedly not open-source, and are often specialized for just one lab task.
This means that the majority of labs in the world, from DIY bio spaces to academic labs and biotech startups, don’t have these robots. And most biologists — the highly educated leading lights of essential scientific research — spend most of their time moving minuscule amounts of liquid around by hand. To do this, they use finely calibrated syringes called micropipettes, which are operated by carefully navigating the pointed tip into a tiny well, pushing down on a plunger with the thumb, and repeating, thousands and thousands of times. To prevent cross-contamination, the pipettes have disposable tips, which are held on by friction and occasionally come loose, messing up calibration and, worst-case, splashing bio-juice all over the lab and ruining everything.
To extend the analogy, most lab workers aren’t living in the mainframe era — they’re stuck hunched over abaci.
Two years ago, Canine had no intention of becoming a biology robot entrepreneur. He describes his background as being in community organizing and political campaigns, and he spent his first two years after college in New York, organizing with Occupy. But then he enrolled in NYU’s Interactive Telecommunications Program, a graduate program that’s become something like the MIT Media Lab’s artier cousin. ITP alumni have created statues that visualize bitcoin transactions, a water bottle that filters every sip, and a concert of new compositions for brand new instruments like the Overtone Harp, “a piano harp mounted vertically and fitted with electro-magnets that stimulate strings in multiple overtone combinations.” Many of these are Kickstarted.
After fiddling around with MakerBots for a while, Canine decided to sign up for the intro synthetic biology class at Genspace, a DIY bio lab in Brooklyn. He had been on a biology track in high school, only to wash out when he placed into the junior-level lab class his freshman year of college. There, back in the fold of wetwork, he had to face an uncomfortable truth: he was awful at pipetting.
“I was like ‘Shit, I didn’t go get a PhD in Microbiology, so I don’t have any lab technique,’” Canine said. But, he thought, who needs technique when you’ve got a robot?
He got in touch with two tech guys on the DIY Bio listserve with more experience than he, Nick Wagner (a software guy) and Chiu Chau (a robotics guy with decades of biology robot experience). OpenTrons was born. They built the robot’s brain around the open-source Raspberry Pi board, and for the robot’s precisely-controlled arm, they used the open-source Smoothieboard. The OT.One is designed to use standard-issue manual pipettes, with a little robotic piston replacing the weary thumb of an unlucky postdoc, and it comes equipped with a webcam so researchers can make sure that protocols are going smoothly (and possibly figure out where things went wrong, if they leave it running overnight).
Soon after the prototype was finished this past spring, in time for Canine’s ITP thesis presentation, Canine, Wagner, and Chau were accepted to HAXLR8R, a Shenzhen-based hardware accelerator. The team set off for China. They lived there all summer, working on making the OT.One into a mass-manufacturable product, and when they got back to the US, they immediately launched their Kickstarter campaign. They haven’t had time to run the tests needed to generate reliable data on their robot’s error rates or accuracy yet, but that hasn’t stopped nearly 200 people from donating over $75,000 to the campaign as of this writing.
Saving grad students from carpal tunnel and letting inept DIY biologists run experiments without years of pipette training is all well and good, but Canine’s vision for OpenTrons goes much farther. With cheap and open-source automation comes standardization, spread out among every lab that’s put down $3000 (or just $2000, if you go for the Kickstarter) for an OT.One. According to some surveys, an astonishingly low percentage — like ten percent low — of experimental results are actually reproducible by another lab following the methods laid out in published studies. It’s unclear why, exactly, this is such a problem, but variations in conditions, protocols, and even pipetting technique from lab to lab are all uncontrolled variables.
Canine believes that widespread automation, combined with the open-source spirit of makers and DIYers and Occupy, might be able to bring that percentage up. And solving that part of the problem will take more than just a cheap robot.
The OT.One is controlled by a browser-based interface called Mix.Bio, which allows users to drag and drop different commands, liquids, and biological components into a protocol that the OT.One then runs. Synthetic biology has always used the metaphor of DNA as a programming language to describe itself, and Mix.Bio puts a user-friendly GUI on top. Mix.Bio also allows users to easily share and download protocols, which brings the whole system back to the MakerBot concept.
“Our whole goal is to be a digital fabrication machine for the life science lab,” Canine says, where “you can email a friend a file, make sure the right filament or material is in the bed, and hit run.”
The OpenTrons system still depends on a lot of manual biology work — the material in the bed of a MakerBot is a bunch of plastic; the material in the bed of the OT.One is a complex and fragile combination of biological building blocks suspended in liquid — but in an ideal world, this kind of automation would allow for fast, trustworthy distributed experimentation. Hundreds of OpenTrons systems around the world could all run the same experiment at the same time, producing orders of magnitude more data than any one lab could on its own, in a relatively short amount of time. With planned additions to the system, like a centrifuge that Canine estimates might be available next summer, these shareable protocols will be able to replicate more and more sophisticated experiments.
Ultimately, the PC analogy is a little unsatisfying for anyone accustomed to thinking in terms of Moore’s Law. Biological processes just take more time than computers. The organisms being manipulated need time to grow, but more importantly, the whole DNA as programming language metaphor is fundamentally misleading. We can’t yet hack directly into life’s command line, just try to fiddle with it over a weak connection and hope the intended changes get through without crashing the whole system. And even then, it takes a couple days (and a lot of pipetting) to even see if anything worked. One of synthetic biology’s biggest success stories comes from UC Berkley, where researchers managed to genetically modify yeast to produce an anti-malarial drug called artemisinin. The lead scientist in the effort, Jay Keasling, estimates that the project took 150 person-years of work to complete.
With that relatively vast complexity and glacial pace in mind, asking Canine about the dangers of someone cooking up a DIY bio version of the 3D-printed gun , some kind of supervirus, or a new strain of antibiotic resistant plague, seemed a little asinine, but he took the question seriously. “We’re right to be afraid of biotech, because it’s powerful, powerful stuff,” he said. “But I think the problem comes not from the technology, but the people who have it, and so far it’s mostly been used to intensify poisonous practices and maintain strangleholds on things like agricultural production.”
Going back to his first metaphor, he continued: “Computers were first developed to drop bombs, it’s not like they were amazing things that everybody made gifs on and smiled about at first. But one thing we’re trying to do is make innovation in biotech accessible for people to make things that aren’t constrained by profit-seeking. The way to make better biotech is not by limiting who can actually make it, but by making it available to everybody.”
All photos by Andrew White/Re:form