German research institute tests Memory-Driven Computing to fight neurodegenerative diseases

As world populations age, the human and economic toll of currently incurable neurodegenerative diseases such as Alzheimer’s grows staggering. DZNE fights these diseases using Big Data analytics, but the limitations of traditional computer systems have been a major bottleneck. Seeking a breakthrough solution, DZNE discovered HPE’s Memory-Driven Computing—and saw unprecedented computational speed improvements that hold new promise in the race against Alzheimer’s.

A global time bomb

As world populations age, diseases of the brain such as Alzheimer’s strike millions of people and cost more than a trillion dollars

Accelerating Alzheimer’s Research – DZNE & Hewlett Packard Enterprise add computational muscle to advance research. Learn more about Memory-Driven Computing.

Featured in this Video: Prof. Pierluigi Nicotera, Scientific Director & CEO, German Center for Neurodegenerative Disease – DZNE.

Prof. Joachim Schulze, MD Professor of Genomics and Immunoregulation – University of Bonn.

Play Video
3:11
Accelerating Alzheimer’s Research

Neurodegenerative diseases—such as Alzheimer’s, Parkinson’s, multiple sclerosis, Lou Gehrig’s disease, and Huntington’s—occur when neurons in the brain and spinal cord deteriorate. The symptoms initially might be slight: problems with coordination or remembering names. But as more and more neurons die, people lose the ability to think clearly, walk independently, and function in the world. Many of the diseases are ultimately fatal.

Because neurodegenerative diseases tend to strike later in life, the incidence is expected to soar as the population ages. The fastest growth in the elderly population is taking place in China, India, south Asia, and the western Pacific.

Dementia—problems with mental functioning—is one of the most debilitating effects of neurodegenerative diseases. The number of people living with dementia is doubling every 20 years, and is expected to exceed 130 million worldwide by 2050. Someone in the world develops dementia every three seconds. Estimates hold that three quarters of people with dementia have not yet been diagnosed. Compounding the problem, by the time a diagnosis is made, it’s too late. The damage is already done.

  • "If we don’t stop the progression in the number of people worldwide developing dementia, by 2050 we would need the entire GDP of the United States today to take care of them."

    Professor Pierluigi Nicotera, Ph.D., Scientific Director and Chairman of the Executive Board, DZNE

The annual global cost of dementia is expected to exceed $1 trillion by 2018. This includes unpaid care by family and others; social care by professional caregivers; and medical care. If global dementia care were a country, it would be the 18th largest economy in the world.

Despite the urgency to find cures, progress has been slow due largely to the complexity of the affected systems. The human brain has 1,000 times the number of neuronal connections as there are stars in our galaxy. Researchers must understand how the brain works, the underlying genetics, cellular and intra-cellular functions, environmental factors that trigger disorders—and how all of these interact over decades.

The amount and types of data such research generates are huge and diverse. In the race to cure neurodegenerative disease, the analytic limitations of traditional computer systems have been a major roadblock.

1B

People worldwide have neurological disorders

24M

Of these have Alzheimer’s

$1T

Global cost of dementia by 2018

Fighting neurodegenerative diseases

German research institution battles brain diseases such as Alzheimer’s and Parkinson’s

  • "We want to use the most modern technology to answer questions such as, ‘Why do we get Alzheimer’s? How can we prevent it?"

    Professor Joachim L. Schultze, Ph.D., Director, PRECISE Platform for Single Cell Genomics and Epigenomics, DZNE

Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE) is a research institution established by the German Federal Ministry of Education and Research to fight neurodegenerative diseases such as Alzheimer’s, Parkinson’s, and multiple sclerosis. DZNE leverages clinical research, population studies, and health care research entailing massive amounts of data. Its researchers are located in nine sites across Germany, and they collaborate closely with universities, university hospitals, and other research partners.

1,000

DZNE employees studying brain diseases

80

Working groups developing new preventive and therapeutic approaches

9

Locations throughout Germany

Unlocking the secrets to early detection

Research projects challenge the limits of technology

 

Researchers know that the disease processes leading to dementia begin decades before symptoms manifest, but they don’t know exactly how. Misfolding proteins? Inflammation? Better understanding is the key to prevention, diagnosis, and treatment.

  • "We’re looking for the little differences that explain why diseases develop in one person and not the other. But we need new technologies to hold all of this data together, and to compare datasets with one another to understand what they mean."

    Professor Pierluigi Nicotera, Ph.D., Scientific Director and Chairman of the Executive Board, DZNE

DZNE is leading a German population study examining up to 30,000 people every three years over their lifetimes to detect changes in gait, sense of smell, and other factors relevant to early detection of Alzheimer’s. Like other DZNE projects, this involves massive amounts of data correlated to detect the small variations that indicate meaningful differences in who will get Alzheimer’s.

Such work requires collaboration—both across geographies and across disciplines including medicine, life sciences, mathematics, physics, informatics, and computer science. Itself spread across Germany, DZNE works with partners globally on research that challenges the limits of technological possibility. Meanwhile as world populations age, neurological diseases are spreading in what some call a senior time bomb.

30

Year population study underway

30K

People tracked

Slow systems hamper progress

Analytic demands surpass the capacity of traditional computer architecture

 

IT limitations have become a major bottleneck in the fight against neurodegenerative diseases. Traditional computing systems are too slow for DZNE’s petabytes of data, multiplicity of data sources, and complex computational pipelines.

DZNE uses information from genomics, brain imaging, and clinical studies, which must be accessed and analyzed securely to protect patient privacy. These huge and diverse datasets were not designed to work together and are often incompatible. Yet researchers want to perform intensive calculations across datasets, such as correlating genetic markers with brain imaging.

  • "We need the computing power to understand these complex diseases on many levels: genomics, brain imaging, patient monitoring over time. Modern medicine will be connected with computation."

    Professor Joachim L. Schultze, Ph.D., Director, PRECISE Platform for Single Cell Genomics and Epigenomics, DZNE

It can take weeks to load the data, and even longer to run calculations on it. Even transmitting the data is infeasible with even the fastest Internet connections. One genomics researcher loads his data onto a hard drive and sends it by truck to DZNE.

DZNE aims to speed up these processes—to stop transporting raw data, and instead analyze it locally, while giving collaboration partners centralized access to results that they can use in their research studies. This requires a new approach to computer architecture.

Memory-Driven Computing fulfills vision

DNZE taps HPE’s pioneering new computer architecture for Big Data

 

DZNE wanted a way to work with genomic data quickly and in a decentralized way, without losing time transmitting data among collaborators or even among its onsite compute tiers. Memory-Driven Computing from Hewlett Packard Enterprise is providing the solution.

Memory-Driven Computing is a new computer architecture that aims to turn the conventional computer architecture inside out. In traditional computer systems, relatively small amounts of memory are tethered to each processor. The resulting inefficiencies limit performance; an estimated 90% of work is devoted to moving information from processor to processor and between tiers of memory and storage.

  • "Memory-Driven Computing provides exactly what we are looking for. By storing a lot of data in memory, we have a much faster system that can accelerate our computational pipelines."

    Professor Joachim L. Schultze, Ph.D., Director, PRECISE Platform for Single Cell Genomics and Epigenomics, DZNE

With Memory-Driven Computing, all processors gain equal access to a pool of shared memory, eliminating the back-and-forth. This delivers unprecedented speed, reliability, and energy efficiency—and ways to harness enormous data sets impossible until now. HPE introduced its Memory-Driven Computing prototype in 2017. Called The Machine, it has 160 terabytes of fast memory, making it the largest single-memory system ever built.

Excited by the promise of Memory-Driven Computing, DZNE leaders selected a particularly challenging use case, tackling an existing algorithm that was already “near to optimal” for pre-processing genomics data. It aimed to see whether making small changes using Memory-Driven Computing techniques could improve a step already running as fast as current technologies could deliver.

The results electrified DZNE.

Computational muscle accelerates research

Memory-Driven Computing brings early diagnosis and treatment closer than ever before

 

DZNE sees Memory-Driven Computing as a breakthrough technology that can unleash researchers’ creative problem-solving and accelerate the search for disease prevention and cures. Having all of the large—and often incompatible datasets available at once in memory clears the computational bottlenecks that have hindered genomics and medical research.

In addition to being faster and more efficient, Memory-Driven Computing is more inherently secure. Instead of sending raw data—such as brain scans—research partners share the results of their calculations, e.g. this brain has a lesion in this location. Insights from the data can be shared with collaborators to advance everyone’s research, but the data stays local; algorithms go to the data instead of the other way around. Security becomes programmatic instead of administrative.

  • "By speeding up our research, Memory-Driven Computing increases the probability that we will find a therapy for Alzheimer’s within a short period of time."

    Professor Pierluigi Nicotera, Ph.D., Scientific Director and Chairman of the Executive Board, DZNE

DZNE and HPE researchers worked together to adapt DNZE’s algorithm for pre-processing genomics data to use Memory-Driven programming techniques.  DZNE has seen a 22 minute process drop to 2.5 minutes, then down to 69 seconds, and now to 13 seconds with some additional coding tweaks. That’s more than a 100x speed improvement after just three months of work! DZNE believes the new architecture can eventually accelerate all of its computational pipelines by 100x.

DZNE is searching for biomarkers that indicate a young person’s probability of developing a neurological disease later in life. It’s searching for cures. With the promise of Memory-Driven Computing, DZNE believes all of this can happen much sooner than ever before possible.

100x

Anticipated increase in analytics speed blasts research bottlenecks

60%

Power reduction cuts research costs

A bridge to Memory-Driven Computing

DZNE used its HPE Integrity Superdome X as a test bed for Memory-Driven Computing programming techniques