IIt’s 1 a.m. on the West Coast of America, but Emerald Cloud Lab, just south of San Francisco, is still busy. Here, more than 100 items of cutting-edge bioscience equipment are on largely unmanned workbenches, 24 hours a day, seven days a week, to conduct experiments for researchers from around the world. I “visit” via camera on a chest-high telepresence robot, as I am driven around the 1,400-square-meter (15,000-square-foot) lab by Emerald CEO Brian Frezza, who is also at home. There are no actual scientists anywhere, just a few staff members in blue coats quietly following instructions from screens on their carts, ensuring the instruments are loaded with reagents and samples.
Cloud labs mean that anyone anywhere can conduct experiments via remote control, using their web browser. Experiments are programmed through a subscription-based online interface – the software then coordinates robots and automated science tools to conduct the experiment and process the data. Friday night is the busiest time of the week in the Emerald, with scientists scheduling experiments while relaxing with their families over the weekend.
There are still some things that robots cannot do, for example lifting giant containers (containers for liquids) or dismantling samples sent by mail, and there are a few tools that cannot be automated. Hence the people in blue coats, who look a bit like Amazon warehouse pickers. It turns out that they are, in fact, former Amazon employees.
Emerald originally hired scientists and lab technicians to help the facility run smoothly, but they were creatively stifled with little work. The poaching of Amazon employees turned out to be an improvement. “We pay them double what they were getting on Amazon to do something more satisfying than stuffing toilet paper in boxes,” says Frieza. “You keep someone’s drug discovery experiment running at full speed.”
To the south in the San Francisco Bay Area are two other cloud labs operated by Strateos. Racks of gleaming life-science gadgets — incubators, mixers, mass spectrometers, PCR machines — sit humming inside large Perspex boxes known as work cells. Arguably the setup is more futuristic than Emerald. Here, reagents and samples travel to the correct working cell on high-tech magnetic conveyor belts and are gently loaded into place by robotic arms. The researchers’ experiences are “non-locational,” in the words of Strateos’ chief operating officer, Mark Celady.
Automation in science is nothing new, especially in fields such as molecular biology, where much experimental work involves the tedious and repetitive transfer of small amounts of liquid from one vial to another. The disruption caused by the pandemic has also encouraged a number of specialized facilities to develop ways to operate their equipment remotely. (UK mighty beam diamond light sourceFor example, a particle accelerator that generates high-energy radiation can now be run to examine matter by users from anywhere in the world.) Outsourcing difficult or time-consuming components of the experimental process is not new either.
But Emerald and Strateos are different – these are the world’s first labs that in theory allow anyone with a laptop and credit card to “pay and play” with the entire detector inventory and a range of devices available in a world-class research facility. The appeal of this approach became apparent during the pandemic, when many researchers were unable to visit their laboratories themselves; The founders of Cloud Labs say this is the future of the life sciences.
The most obvious benefit is throughput: Researchers can run many experiments simultaneously and queue them up to run overnight or while they do other things. “Our pro users, they will be doing 10 scientists in a conventional lab,” Frezza says. “They’re going to shoot ridiculous numbers.”
There is no time to spend preparing and disassembling equipment, cleaning, maintaining, repairing or replenishing inventory. Arctoris, a remote drug discovery lab in Oxfordshire, says its platform has completed projects for drug companies in 24 hours that would take at least a week in a traditional environment. Instead of sucking on for hours each day, researchers can spend more time thinking, reading, and analyzing the results with colleagues.
Scientists at Carnegie Mellon University in Pittsburgh were so impressed with what staff and students at Emerald Cloud Lab could do — one researcher was able to recreate years of his PhD experiments in a matter of weeks — that they recently asked the company to build another experiment, just for them.
Because the cost of accessing a cloud lab for a year is often less than the price of one piece of high-end lab equipment, Carnegie Mellon dean Rebecca Durge says. “I am not only interested in changing the science at Carnegie Mellon University. I am interested in changing the science process around the world,” she says of the new facility in Pittsburgh. “We all have colleagues in under-resourced places who just can’t do the science they can do because they don’t have enough money. So with internet connectivity and access to a cloud lab, that is a game changer.”
Doerge, a statistician turned science official, is also excited about removing variance and human error from experimental work. There will be no scientists at the new 1,500 square meter (16,000 square foot) site, just six technicians assisting the venue 24 hours a day. “People still go to wet labs and they still stand there and make mistakes. I don’t think everything can be automated in science, I’m not saying that. I’m just saying that repetitive things, once you learn them, you don’t need to stand there and do it over and over again. “.
Scientists like Doerge believe the accuracy of remotely operated labs can help fix what has come to be known as ‘science’.Reproduction crisis– The disturbing revelation that the results of published research groups cannot be replicated when different groups of scientists follow exactly the same methods. Entering an experiment in a browser to be performed by robots forces researchers to translate the minute details of each step into unambiguous code. For example, What was once described in a scientific paper as “mixing samples” becomes detailed computer instructions for a particular machine to mix a certain number of revolutions per minute for a given period. Other factors that can affect the result, such as the ambient temperature at the time, are recorded in the metadata. .
As Doerge encouraged more and more research – and even teaching – at Carnegie Mellon to be taken to remote laboratories, not all of her colleagues were supportive of her. Many scientists believe that it is working alongside colleagues on the bench and the sights and sounds of experimentation that help generate exciting ideas and happy accidents. Others have concerns about the quality of data being produced in labs that have never set foot in. “If I don’t see it with my own eyes, it doesn’t exist”—I heard it from some senior faculty members,” says Dorge. “It’s definitely a mind shift.”
Some experts believe that making advanced laboratories so easily accessible is a potential threat to biosecurity or bioterrorism. In theory, small groups or even individuals without research experience could use the cloud lab to start performing complex biological experiments. “The labs say they only work with trusted partners, but of course they are very keen to open up their market,” says Dr Philippa Lentzos, an expert in biohazards and biosecurity at Kings College London. “Although we have to remember that most people come from a good place, there are some pretty crazy people out there as well. The barriers definitely go down if you want to do something intentionally harmful.”
Cloud Labs says it reviews all scheduled trials and has systems in place to report or reject any that appear illegal or dangerous. Plus, they say, the complete digitization of everything that happens in the lab makes it easier to record and monitor what people are doing than in a traditional lab.
Paul Fremont, co-founder of the UK Center for Innovation and Knowledge for Synthetic Biology, helped develop several highly automated laboratories in the UK, including a robotic platform that was able to perform more than 1,000 Covid-19 tests per day early in the pandemic. He’s not sure that remotely operated labs are “mature” enough to replicate what’s available to scientists who set up their own robotic equipment. “I like this concept and I think that’s the way science goes. It would work if we had all the necessary protocols and workflows that a biologist would need, but I think that’s not currently available for the level of complexity and detail that one needs.”
Freemont also has concerns about scientists not really understanding or dealing with the software or hardware that generates their data. “You have to have the next generation of scientists who understand how to build all this infrastructure themselves and how to work with it – you have to have some hands-on experience for sure. The prospect of a few big private labs or companies monopolizing that understanding – I don’t think it would be very healthy. “.
Despite these concerns, the uptake of cloud science is growing. Emerald is expanding its capacity to keep up with demand, mostly from pharmaceutical companies and biotech startups. Strateos is working with US research agency Darpa to study in detail how its facilities can improve the reproducibility and efficiency of past experiments, and the company is also licensing its software so other organizations can transform its facilities.
In the future, cloud labs may decide which experiments to do themselves. as Google DeepMind recently provenMachine learning tools can now gobble up decades of data and reveal answers to questions that may take scientists many years to solve through physical investigation. Pharmaceutical companies are increasingly using these tools to simulate molecular interactions in their search for new drugs. Data generated through cloud labs — which turns biology into information technology — will only make these tools more powerful. Combining all of these technologies could one day lead to systems that can develop and physically test theories without human intervention.
Already, some advanced Emerald Cloud Lab users have developed algorithms that adjust the parameters or direction of the next experiment based on their data analysis. “It’s kind of wild, and very futuristic,” says Frieza.
All of this means that scientists are the latest profession asking what it means to move towards automation and artificial intelligence for the future. Could traditional research scientists ever find themselves out of work? Unlikely – after all, we will always need to prioritize the questions that need to be answered and develop new ways to answer them. But the days of sitting on a bench in a white coat and gloves by the flame of a Bunsen burner may soon be a thing of the past — the era of the robotic researcher is coming.