Science labs in the cloud: Champagne discoveries, beer budget

Science is hard. Repeating other labs' findings is even harder.

That's particularly true in the life sciences, where biologists and researchers manually run sophisticated experiments fraught with variables such as water purity, humidity or subtle differences in equipment.

"Biologists do almost everything by hand, and there's a lot of variation in experiments from lab to lab," said Ethan Perlstein, founder and CEO of Perlstein Lab, a San Francisco-based biotech startup working to discover drugs to fight rare genetic diseases.

"I want to be able to do science in a way that's auditable and scalable, and the data is repeatable, pristine and good."

Emerald Therapeutics thinks its Emerald Cloud Lab, dubbed ECL-1, could deliver just what Perlstein ordered. That's because ECL-1 lets scientists remotely conduct more than 40 different standard experiment types on $3 million worth of lab equipment, including DNA synthesizers, advanced microscopes and mass spectrometers for separating and measuring the weight of charged particles. All similar equipment in the 15,000-square-foot facility is perfectly identical, down to the type and length of tubing and pipes. Automated robots perform the tests, removing the human element from the equation.

Other startups offer similar science labs in the cloud. DNAnexus provides tools to store, manage and analyze the enormous amount of data generated by genome sequencing. Transcriptic also touts its automated robotics and controlled lab environment, but it handles fewer than 20 experiments, compared with ECL's 40.

In effect, Emerald and companies like it aim to become the biosciences version of Amazon Web Services. Just as AWS pared the cost of developing commercial software, labs in the cloud may change the economics of life sciences exploration.

The average cost per experiment run by ECL-1, for example, is $25.

"We want to be the Amazon Prime of experiments, starting within 48 hours of the order," Brian Frezza, Emerald Therapeutics' co-founder and co-CEO, said from the company's offices in South San Francisco, Calif.

Instead of leasing computer servers by the hour, as Amazon Web Services does, Emerald and its rival startups rent out robots and precisely calibrated instrumentation -- with the added benefit of providing an auditable data trail.

"Having robots handle experiments on exactly the same equipment, in exactly the same environment, will always be superior to two pairs of hands," said Perlstein. "It's just objectively better to have fully auditable and automated experiments."

The ramifications could be enormous. Consider: In 2011, researchers at German pharmaceutical giant Bayer HealthCare reported they could replicate only one-third of 67 key studies published in top science journals. The next year Amgen's former head of global cancer research told Nature magazine that his team failed to reproduce the findings of 47 out of 53 "landmark" studies the pharmaceutical company wanted to use in its drug development work. Variables in equipment and environment could have contributed to their issues.

Having reproducible results isn't the only potential boon from labs in the cloud. There's also a financial advantage, since modern labs buy equipment costing anywhere from $100,000 to $240,000 per piece, which they use for perhaps 10 percent of their research.

"This sort of experiment as a service is interesting because it helps avoid the huge setup costs of buying the equipment as well as finding someone skilled enough to understand and configure everything -- all before you can even run your first experiment," said Frank Gillett, an analyst with Forrester Research. "Now you can reduce the time it takes to complete the experiment and reduce the cost of the experiment itself."

Stephen Wolfram, developer of the Mathematica computational software and the Wolfram programming language (and one of the youngest recipients of the MacArthur Fellowship), sees an even bigger benefit: labs like ECL-1 let scientists explore the biggest questions.

Wolfram calls it a "top down" versus "bottom up" approach to scientific discovery. A top-down approach lets researchers design whatever experiments will "get at some larger truth about biology." In a bottom-up approach, however, scientists decide what to explore based on the resources they have at their disposal. While both are valid, the top-down way of thinking leads to big discoveries faster. Wolfram is also an adviser to Emerald Therapeutics.

With ECL-1, scientists "can be running any experiment remotely within 24 hours, all without having to invest months of time, dedicating large number of staff members, and obtaining millions of dollars in fixed cost equipment to bring these experiments online," Wolfram said in an emailed statement.

"Life scientists can go back to a focus on how to answer the largest remaining questions in chemistry and biology rather then spend all of their time focused on becoming professional fundraisers and managers of large organizations," Wolfram said.

Emerald expects to offer more than 100 experiments -- or 100 percent of all life science tests -- within the next 18 months. "We won't consider ourselves in full production until we offer 100 percent of all the available lab tests," said DJ Kleinbaum, Emerald's other co-founder and co-CEO. Best friends since grade school, Kleinbaum holds a Ph.D. in organic chemistry from Stanford University; Frezza earned his Ph.D. in chemical biology from The Scripps Research Institute. They founded Emerald in 2010.

For his part, Perlstein is happy to have access to ECL's available tests, conducted with robotic precision. "This is the beginning of the transformation to replicable scientific methods," said Perlstein. "Science isn't science until it's reproduced."