Stanford Takes Biology Experimentation to Cloud

• By Dian Schaffhauser
• 04/22/15

A project going on at Stanford University offers the promise of allowing online students and others with an interest to perform live biology experiments over the Internet. This "interactive cloud experimentation," according to the researchers, does away with the "complex logistics of experimentation" and allows people to focus on the data analytics.

The goal was to develop an architecture for cloud experimentation for biology that could scale to large numbers of users and diverse applications "in a cost- and logistics-effective manner." The initiative uses a robotic biology cloud lab to carry out remote-control experiments. The idea was to set up a system that would allow multiple users to access and run experiments anytime without having to "book a time slot."

As described in the paper, "Interactive Cloud Experimentation for Biology: An Online Education Case Study," the experiments are executed in multiple "biotic processing units" or BPUs. These are defined as "automated hardware that houses a specific set of biological specimens in one or more isolated compartments" or "wells." Each well is shared among one or more users doing a given experiment. Users use a Web application to provide small sequences of instructions in "blocks." Up until a block is performed, it can be edited, deleted or added to other blocks. The BPU aggregates blocks from users that pertain to a certain time and "interleave them" for optimal batch execution. Each BPU includes a microcontroller for interpreting instructions and acquiring the output from the experiment and sending the resulting data to a storage server.

To test the system, the researchers incorporated a BPU as a lab component in a graduate level theory class. Zahid Hossain, a Stanford doctoral student, constructed a prototype BPU by using Lego Mindstorms to build a liquid-handling robot. This robot traveled over a flatbed photo scanner that held petri dishes containing the slime mold Physarum polycephalum, which "eats" oatmeal.

The researchers set up three of these BPUs, each with six petri dishes and housed in a server rack. Students in the course used remote-control interfaces on their smartphones to order the robot to drop oatmeal onto specific petri dishes. The software allowed them to choose different droplet patterns. The scanner recorded how the mold followed each trail of oatmeal dots by "sniffing out" chemical cues in the petri dishes.

While the prototype supported 18 users and allowed the researchers to assess the viability of cloud labs, Hossain said he would like to see "advanced BPUs supporting many different types of experiments and thousands of different users."

As the researchers noted in their paper, the project led to several "practical lessons" for future experimentation:

• The creation and use of cloud systems for biology requires expertise in several areas, including biology, mechatronics, database, Web interfaces and education;
• Building and maintaining the BPU and its biological content is challenging because the biological specimens "must be stable and responsive to defined stimuli over a long time";
• Once they try remote experimentation, users will "likely request additional features; the researchers recommend that the user interface provide a means for people to submit those requests'; and
• Critical features of the user interface should be tested out by users in advance; the ones that aren't critical can be "beta tested during the actual usage."

"Biotechnology today is very similar to where computing technology used to be," said Ingmar Riedel-Kruse, an assistant professor of bioengineering involved in the project. "Biological labs are housed in big buildings and the technology is hard to access. But we are changing that. We are enabling people to interact with biological materials and perform experiments the way they interact with computers today."

This cloud lab research is being presented at the Computer Human Interactions (CHI) conference in Seoul, South Korea.