DREAM Challenge Tackles Alzheimer’s Biomarkers

This summer, aficionados of big data can turn their attention to AD biomarkers. The Alzheimer’s Disease Big Data DREAM Challenge No. 1 is the first of several planned projects to try to accelerate Alzheimer’s research by uniting computer programmers, geneticists, clinicians, and other specialists in a common goal, said Thea Norman, coordinator of the DREAM (Dialogue on Reverse Engineering Assessment and Methods) contests, which are now in their ninth year. These experts will work together, using single-nucleotide polymorphism data, medical images, and clinical information provided by the challenge, to write algorithms that help diagnose Alzheimer’s and predict its course. The teams that construct the most accurate algorithms will have the opportunity to co-author a paper for submission to Nature Neuroscience. The challenge officially opened June 2.

Past DREAM challenges have tackled cancer, amyotrophic lateral sclerosis (see Nov 2012 news story), and other problems that take a concerted effort to solve. The AD challenge differs from most other Alzheimer’s biomarker efforts in its multidisciplinary nature, organizers said. “I would like to see a community of scientists with diverse skills start working together, and I would like to see some interesting and encouraging predictions coming out of the challenge,” said Norman, who works at project partner Sage Bionetworks in Seattle. More than 250 people have registered to participate, Norman said, and she expects about 500 registrants by the time the challenge closes in September.

Contestants will address three questions chosen by a scientific advisory board as being high-priority and amenable to complex number crunching. The organizers will provide data from the Alzheimer’s Disease Neuroimaging Initiative, The AddNeuroMed Study, and Rush University Medical Center’s Religious Orders Study and Memory and Aging Project. For each question, competitors will have one “training” data set, which includes the “answers” to help develop their algorithms. The organizers have designated other test data sets.

Subchallenge No. 1 concerns cognitive decline. Competitors will start with baseline mini-mental state exam (MMSE) scores, genotypes, and clinical and demographic data. From there, they will design tools to predict what the subjects’ cognitive scores will be two years later. These predictions would help patients and clinicians plan for the future, and help researchers design efficient trials by selecting subjects who are on a fast disease trajectory, said John “Keoni” Kauwe of Brigham Young University in Provo, Utah, the scientific leader of the AD No. 1 Challenge.

Subchallenge No. 2 asks participants to write an algorithm to identify people who stay mentally sharp despite having low Aβ42 in their cerebrospinal fluid, which indicates risk of Alzheimer’s dementia. Competitors will start with demographic data, genotypes, and baseline MMSE scores, but not baseline Aβ levels, as input. The goal is to figure out who has low CSF Aβ yet remains cognitively normal. Successful algorithms would identify genetic or other factors that make those people resistant to dementia. Those factors might suggest ways to stave off AD in others, Kauwe suggested.

Subchallenge No. 3 requires contestants to use brain magnetic resonance images to diagnose Alzheimer’s. Algorithm input will include MRIs plus genetic and demographic data. The programs should identify who has Alzheimer’s or mild cognitive impairment or neither, as well as predict participants’ MMSE scores. Imaging data are complex, Kauwe noted, and there are many ways to evaluate them that have not yet been explored. Participants can submit algorithms throughout the summer and see how they measure up against other teams’ predictions on a leader board.

Norman and colleagues see the challenge as the jumping-off point for a larger effort linking experts who normally would not work together to understand AD. “These challenges represent the beginning of research initiatives, not the end,” said Norman, who hopes competitors will continue to collaborate after the final bell sounds on the 2014 contest.

The Global CEO Initiative on Alzheimer’s Disease, one of the contest partners, will run a meeting later this month to discuss how big-data studies can contribute to Alzheimer’s research generally, said George Vradenburg, who convened the initiative. Future DREAM challenges addressing AD may incorporate exome and whole-genome sequences to the single-nucleotide polymorphisms included in the current project.—Amber Dance

Comments

No Available Comments

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. Contest Winners Offer Solutions for Tracking ALS

External Citations

  1. Alzheimer’s Disease Big Data DREAM Challenge

Further Reading

Papers

  1. Paquerault S. Battle against Alzheimer's disease: the scope and potential value of magnetic resonance imaging biomarkers. Acad Radiol. 2012 May;19(5):509-11. PubMed.
  2. Chiu MJ, Yang SY, Horng HE, Yang CC, Chen TF, Chieh JJ, Chen HH, Chen TC, Ho CS, Chang SF, Liu HC, Hong CY, Yang HC. Combined plasma biomarkers for diagnosing mild cognition impairment and Alzheimer's disease. ACS Chem Neurosci. 2013 Dec 18;4(12):1530-6. Epub 2013 Oct 23 PubMed.
  3. Czapski GA, Maruszak A, Styczynska M, Zekanowski C, Safranow K, Strosznajder JB. Association between plasma biomarkers, CDK5 polymorphism and the risk of Alzheimer's disease. Acta Neurobiol Exp (Wars). 2012;72(4):397-411. PubMed.

Annotate

To make an annotation you must Login or Register.