Cloud computing is rapidly catching on for businesses and consumer applications such as Gmail and Netflix. But scientists are discovering that the cloud can make their research a lot easier as well. In a news feature for Nature Methods, reporter Vivien Marx looked at how genomic research fits into the cloud, offering new opportunities for collaboration and data management. Globus Genomics, the CI's cloud-based analysis platform for gene sequence data, is heavily featured in the article, as are the CI's Paul Davé and Ravi Madduri.
The team tested the platform with the lab of William Dobyns, at Seattle Children's Hospital, who studies the genetics of brain malformation. His patient samples were sequenced at the Broad Institute and by a service offered through PerkinElmer. Previously, the lab received raw data on hard drives and then struggled with storage and analysis. “To run one of their exome analysis pipelines might take them 20 or 30 hours,” Davé says. Running the analyses one after the next could mean many months of computational time.
In their test, the Globus Genomics team first streamlined data movement so that the data ran securely from sequencing facilities to the analysis pipeline on the cloud as they were generated. This way, says Davé, “they wouldn't queue up raw data from 20 patients and then put it all on one hard disk and send it out.”
The Globus Genomics team members have engineered analysis to scale to the needs of a given pipeline and sample size. And they have optimized genomics analysis tools for their cloud, too. “We take a tool and see: how many processors is it using?” says Madduri. They also look at how much memory tools require and how to match the software settings to those of Amazon's machines. This profiling helps with analyzing multiple data sets in parallel on multiple cloud-based processors, which saves researchers time. “It used to take them 20 hours to get through one exome; they can run 20 exomes in the same amount of time,” says Davé, referring to the test with the Dobyns lab.