The Mind and The Machine: How Memory-Driven Computing is Contributing to the Cure for Alzheimer’s

by Curt Hopkins, Managing Editor, Hewlett Packard Labs
Oct 11, 2017 12:15 PM ET

HPE | Behind the Scenes @ Labs

When the German Center for Neurodegenerative Disease (DZNE) ran one component of its current gene assembly pipeline on a supercomputer with a traditional computing architecture in April, it took 22 minutes to run the data. The last time they used The Machine – HPE’s Memory-Driven Computing system – in early July, it took 36 seconds.

That’s breathtaking in any context. But DZNE researches the causation and treatment of Alzheimer’s and other cognitive diseases, an increasingly serious issue in our aging world.

“According to some estimates if we do not stop the progression of dementia in the population by 2050, we would need the entire gross domestic product of the U.S. to be able to take care of people affected,” explained Dr. Pierluigi Nicotera in an interview at HPE’s Discover conference.

DZNE’s capacity to run its data quickly and accurately may make the difference between an economically disastrous flood of dementia and a cure.

“Because the disease begins about 20 years before the symptoms we need early diagnosis,” said Nicotera. “We need new therapies, we need to improve how we do science, and of course we have to collect a lot of data, a huge amount of data.”

DZNE already handles a massive amount of sensitive data. Their keystone study will follow 30,000 subjects for several decades and they have begun to add data-intensive genomic data to it.

Their recent addition of genomic data to the study has made their data demands even greater. The pipeline they ran so successfully on The Machine was an exercise in genomic assembly, which involves matching bits of automatically sequenced genome into a total picture of a person’s genetic life.

DZNE has never been able to work with such a huge amount of data at one time before. Because such a large shared memory pool is now available to the researchers, they can tackle the data at once, in memory, rather than approach it in pieces. This may allow them to derive new connections that eluded them before., and with persistence, connections may lead to conclusions about cause and treatment. 

Nicotera’s hope, like that of colleagues like Dr. Joachim Schultze, is that Memory-Driven Computing will provide them with the tool their data demands.