Alzheimer Drugs Keep Failing, Can Big Data Help?

This week another pharmaceutical company announced it had given up on developing an Alzheimer’s drug, and the news was not a surprise. It follows announcements just this past May and February of other abandoned drug candidates to slow the disease, and dozens more over the past 20 years.

Many of the failed drug candidates have something in common: They target amyloid, which accumulates in the brain of people who go on to develop Alzheimer’s. Have drug developers been focusing on the wrong target?

“Amyloid is most certainly implicated in Alzheimer’s, but the disease is far more complex than we’ve appreciated,” says Philip L. De Jager, MD, PhD, a neurologist at Columbia University’s Vagelos College of Physicians and Surgeons.

“It’s clear that we need to find completely new strategies to prevent or at least slow the progression of the disease, but we’re just now developing the tools to understand the disease’s complexity.”

“Big data” approaches like network analysis have been used to unravel the complex molecular pathways that govern cancer, but their use in Alzheimer’s research has lagged behind.

“One advantage with cancer is that tissue samples are usually accessible,” De Jager says. “In neurological diseases, we haven’t had enough brain tissue from people who were tracked carefully as they developed dementia to do these sorts of analyses.”

De Jager’s recent work suggests that Alzheimer’s is now ready for big data studies.

Over the past 20 years, De Jager’s colleague David Bennett at Rush University Medical Center has been building up a repository suitable for such data mining. The repository now includes years of data from more than 3,000 people, including how cognition changed over time in each individual and samples of brain tissue from the more than 1,200 participants who are now deceased.

Recently, De Jager, Bennett, and colleagues at the University of British Columbia and Brigham and Women’s Hospital began to tap into the resource. Their effort is part of the Accelerating Medicines Partnership for Alzheimer’s Disease (AMP-AD), a collaboration among the NIH, pharmaceutical companies, and nonprofit organizations to transform the way new Alzheimer’s treatments are developed. De Jager and Bennett were one of the first three groups awarded funding from AMP to use big data techniques to identify new drug targets.

De Jager’s team started by sequencing more than 12,000 genes in each individual brain to determine their level of expression. Rather than focus on individual genes, the researchers used a type of analysis that creates a network of genes that work together to hasten cognitive decline or accumulation of brain pathologies.

With this network diagram in hand, they then looked for the small number of genes in the network that appear to drive the changes seen in the study’s participants. Such “driver genes” are potentially excellent targets for new Alzheimer drugs, because as the network’s nodal points, turning them on or off may have the greatest therapeutic effect. One can think of them as the switches in a complicated circuit that open or close a whole series of downstream events.

"Capturing all of these data allowed us to look for patterns related to the disease without making any assumptions about what was important,” De Jager says.

The researchers recently reported results from this analysis in Nature Neuroscience. They describe 11 groups of genes or modules that appear to directly affect cognitive decline, amyloid, and other Alzheimer’s-related traits.

The team focused attention on one module that seemed most related to cognitive decline and among its 390 genes identified two bona fide driver genes that, when active, increase the production of amyloid. One gene had never been linked to Alzheimer’s. By silencing either one, amyloid levels in the cells were significantly reduced.

These genes may be good therapeutic targets, says De Jager, though more work is needed to confirm their role in disease and to understand how they function.

Many other therapeutic targets within the data await discovery (the researchers only validated the role of one of the 11 modules involved in the disease). In an effort to accelerate Alzheimer’s research worldwide, all of the data and the resulting network models have been made publicly available to encourage independent investigations by other researchers.

“What’s really significant about our study is that it demonstrates how big data and advanced analytic techniques can be used to identify genes that are important in Alzheimer’s disease and are excellent new targets for drug development,” De Jager says.


Dr. De Jager is the Weill-Granet Professor of Neurology in the Department of Neurology, the Center for Translational & Computational Neuroimmunology, the Taub Institute for Research on Alzheimer's Disease and the Aging Brain, and the Columbia Precision Medicine Initiative at Columbia University Vagelos College of Physicians and Surgeons.

This work was supported by the National Institutes of Health (grants U01AG046152, R01AG036836, P30AG10161, R01AG015819, R01AG017917, and R01AG036547) and was done as part of the National Institute of Aging’s Accelerating Medicines Partnership for AD (AMP-AD).

Other authors: Sara Mostafavi (University of British Columbia and Canadian Institute for Advanced Research); Chris Gaiteri, Shinya Tasaki, Vitalina Komashko, Lei Yu, Julie A. Schneider & David A. Bennett (Rush University Medical Center); Sarah E. Sullivan, Robert Smith & Tracy L. Young-Pearse (Brigham and Women's Hospital); Charles C. White, Jishu Xu, Mariko Taga, Hans-Ulrich Klein, Cristin McCabe, Elizabeth M. Bradshaw, David E. Root, Aviv Regev & Lori B. Chibnik (Broad Institute); Mariko Taga, Hans-Ulrich Klein, Elizabeth M. Bradshaw & Philip L. De Jager (Columbia University Irving Medical Center); Ellis Patrick (University of Sydney); Lori B. Chibnik & Tracy L. Young-Pearse (Harvard Medical School); Lori B. Chibnik (Harvard T.H. Chan School of Public Health).

The authors declare no competing interests.