Transcriptic Aims to Make the Biology Lab Programmable

August 13, 2014


By Bio-IT World Staff 

August 13, 2014 | When he was a student at Duke University, Max Hodak was assigned to work on a daunting robotics interface. As a research assistant in the lab of famed neuroengineer Miguel Nicolelis, Hodak had two problems that few programmers have to contend with. First, many of the robots he worked with had devilishly complex feedback systems, including hands with primitive nerve signals that could distinguish objects by touch.

Second, the operators were monkeys.

Specifically, they were rhesus monkeys, and Hodak’s job was to help build interfaces directly between their brains and the robots that acted somewhat like extra prosthetic limbs. By thought alone, the monkeys learned to operate these machines to perform delicate tasks with distant objects, for which they were rewarded with fruit juice. The Nicolelis lab has racked up a remarkable string of successes, with both rhesus monkeys and rats, translating very noisy electroencephalography signals into commands for increasingly complex machines.

Yet when it came time to work in the wet lab, Hodak found that the monkeys had access to better equipment than he did. “In biology labs, you spend all this time just pipetting small volumes of liquid around all day long, or waiting on a sample to come out of an incubator, or a centrifuge to finish,” he tells Bio-IT World. “You feel like a robot.”

At a time when monkeys are exerting telepathic control over mechanical limbs, most biologists still don’t have access to robotics for even the simplest procedures. At large biotech companies like Genentech and Amgen, in-house robotics labs can turn over repetitive, labor-intensive work like high-throughput sequencing day in and day out, but these systems are far too expensive for any but the most lavishly funded labs to install. Even where robots are already in place, their routines are inflexible and can’t be turned to new procedures without a lot of hands-on help from technicians. The nature of the lab turns working scientists into, well, just more pieces of equipment, following rote tasks instead of contributing creatively to their projects.

In February 2012, while he was still a student at Duke, Hodak founded his second company, Transcriptic, to address this gap. In its lab space in Menlo Park, California, Transcriptic operates a fleet of robotics systems designed to work through the steps of complex lab procedures, with the ultimate goal of making most biology experiments possible without human intervention. The company began offering a few simple services, like genotyping mouse samples, in the fall of 2013, but it only released its major product this summer. With the launch of the cloud-based Transcriptic Platform in July, customers can now design their own lab protocols online, and let the robots at Transcriptic do the rest.

A Modular Lab 

Transcriptic is not the only company in the world letting biologists outsource their basic tasks to a robotics lab; in fact, it’s not even the only such company in Menlo Park. The nearby Emerald Cloud Laboratory opened with a very similar service this July, and is now offering applications like Western blots and MALDI mass spectroscopy. (See, “Robots for Hire: Emerald Launches Robotic Laboratories for Life Sciences.”)

But Transcriptic’s vision is not just to offer a menu list of tests that its robots are preset to perform. Hodak wants to open the lab up to custom procedures at whatever level of complexity the customer needs. “The key to this business is that you can cover a huge swathe of biology with only ten to fifteen different devices,” he says. “These totally different areas of research all use the same processes: incubation, centrifugation, liquid handling, plate reading.” By building a lab where samples can pass between these instruments in any order, Transcriptic lets its users go “off script” and create new applications.

While Transcriptic has rarely let journalists peek at its lab space, Hodak describes it as a set of modular units, which can be scaled out horizontally if demand outpaces the available equipment. The commercial instruments in those units are combined with custom-made robots to operate them, as well as to transport samples between devices according to commands from the Transcriptic Platform.

Transcriptic arm 

A robotic arm operating in one of the Transcriptic lab's work cells. Image credit: Transcriptic 

The idea is to create a “programming language” for the life sciences, where each process in the lab is one step in a computer protocol. “When we started this, the joke was that principal investigators already have a high-level programming language for biology,” says Hodak. “You tell your grad students what to do, and they do it.” But despite reservations that PIs would want their students practicing basic lab procedures, Hodak has found that many are eager to free their PhDs from grunt work and let them focus their energy on interpretation and designing new experiments. In fact, most of Transcriptic’s early customers have come from academia, including users at UC Davis, Stanford, and Harvard.

Today, the majority of customers are using Transcriptic much like a Contract Research Organization (CRO), ordering package services like antibody screening or nucleic acid extraction. “That’s the way the market thinks,” says Hodak, who plans to continue adding new packages, including tissue cultures, over the next few months.

In the long term, however, he’s hoping to change the life science industry’s basic ideas about contracting out lab work. As a handful of early adopters graduate to using Transcriptic’s API to build their own experiments, Hodak believes that his robotics lab will begin to open up new applications.

Speaking Transcriptic 

The Transcriptic Platform offers two different APIs. The high-level API is similar to a more customary menu service: users choose the overall process they want to perform — say, DNA synthesis — and then specify any important variables, like the nucleotide sequence they want synthesized. The platform will understand what steps go into that process, and once a contract has been agreed on, the lab whirs into motion.

The low-level API, however, is the truly unique part of Transcriptic’s service. Here, users specify operations like “pipette” or “thermocycle,” each of which is minutely controlled: what volumes are pipetted between which wells, the temperature and duration of thermocycling. Writing instructions is more complicated at this level, and users have to know a major programming language such as Ruby or Python, but the reward, says Hodak, is a set of applications that no other service can offer.

“If you just want to outsource an assay, there are a lot of CROs that can do that,” says Hodak. “But when it becomes programmable, that opens up a lot of possibilities that aren’t available today.” He gives the example of running a new PCR reaction, and not being sure what the correct thermocycling procedure should be. Normally, he says, “if you run a PCR reaction, and it fails, you tweak the parameters applying rules of thumb that the post-docs and lab techs know.

“But with a system like this, you can gather all the data from all the reactions you’ve ever run, and then build models to predict what parameters you should be choosing.” By tracking past runs, it becomes possible to take those rules of thumb out of the hands of lab techs, and let the platform choose the best procedure in a data-driven fashion. “Then, when a reaction fails, you can automatically pick a new set of parameters and try it again. You get a closed-loop feedback cycle with no human interaction.”

This service won’t necessarily be attractive to the majority of biologists, but for the growing field of bioinformaticians, and computer-savvy biotech startups, it could be an empowering way to connect lab procedures directly to the data they produce. While Hodak says that most of Transcriptic’s current customers don’t program, his company has attracted a few users who are primarily computational scientists with little lab experience. “They want to test mathematical models [of biological systems] without going into the lab,” he says. “And that’s perfect for us.”

The Robotics Marketplace 

Although Hodak hopes these custom protocols could one day become his primary business, there are reasons to think robotics labs will offer advantages over CROs even with simple package services.

One of Transcriptic’s first customers is Jack Lin, the founder and CEO of Anvil Biosciences, which is trying to bring gene therapy platforms through preclinical validation. Anvil has early leads in treating rare ocular and neuroinflammatory diseases, but developing them has required several procedures that are very demanding for a small, five-person biotech that started out sharing lab space with three other companies.

Lin stumbled upon Transcriptic by accident, when he was delivering a rent check in the Menlo Park technology park where Anvil had set up shop. Transcriptic was located in the same facility, and Lin noticed their sign as he was passing by. “I thought that was pretty interesting, because we were then at a point where we needed to clone some genes into our plasmids,” he remembers. Rather than purchase the large, expensive equipment involved in bacterial cloning, Lin decided to see if Transcriptic was open for business.

“I just walked in,” he says. “They were surprised that anyone would be contacting them that early, but happy to have me.” The price turned out to be competitive with CROs, and the turnaround time was notably faster. Since that first order in the fall of 2013, Anvil has regularly contracted out to Transcriptic to create libraries of BAC clones. “The communication with them [has been] better than working with a CRO,” Lin adds, “in that they will show you graphically, using DNA sequence viewers and cloning viewers, how the cloning is going to go. So you have peace of mind as to what product you receive.”

Anvil expanded into its own lab space this summer, but Lin continues to work with Transcriptic for cloning, and doesn’t expect his own company’s growth to change that anytime soon. “I specifically still have not gotten anything for molecular cloning and bacterial workflow,” he says. “It’s more of a rote procedure, so I would rather focus our energy on more important experiments, and thinking about more important scientific questions, rather than something that should be routine.”

As the technology matures, robotics labs will have a number of opportunities to outpace traditional CROs. Experiments could be made more reproducible, with the equipment performing each task precisely the same way each time, to a degree of precision that human hands can’t match. And aside from the initial capital costs of the handling equipment, a robotics lab can operate much leaner than a fully staffed facility: Transcriptic employs just two full-time technicians, whose primary role is to work out bugs and help transition new tasks over to the machines.

Hodak anticipates the number of robotics labs will keep growing. “The automation we’re using is not fundamentally different from the automation that any big drug company has,” he says. “Opening it up as a service is new, but it looks obvious in retrospect.” He sees the expansion of these services as healthy for the industry, and not especially worrying for his own business.

“The CRO market has always been very fragmented,” he points out. “I don’t see any indication this is going to become winner-take-all.” At the same time, his company has a chance to set itself apart through its API. By giving customers fine control over their remote lab space, Transcriptic could lower the barriers for a handful of post-docs with a novel idea to try their hand in the life sciences market, in much the same way that cloud services have encouraged the growth of officeless software companies.

Lin certainly believes that Transcriptic offered that kind of freedom to his own startup. “It’s really helped us in terms of getting up and running,” he says. “I’m hoping they bring greater accessibility to the field, for starry-eyed entrepreneurs like myself.”