12 Jul 2017

Biostatisticians, mathematicians, machine learning aficionados, listen up. Scientists are offering £30,000 to the teams who can identify what data, processing pipelines, and quantitative algorithms best predict Alzheimer’s progression. The object of the challenge is to predict who will develop clinical, cognitive, and magnetic resonance imaging (MRI) signs of disease in a short enough timeframe to carry out a clinical trial.

“One of the key reasons why clinical trials in dementia treatments haven’t worked is the heterogeneity of cohorts that go into those trials,” said Daniel Alexander, University College, London. Alexander is principal investigator of the European Progression of Neurological Disease (EuroPOND) initiative, which is collaborating with the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to put on the challenge. “If we can better identify groups of subjects who are likely to need, and can still respond to, a treatment, then hopefully effective treatments wouldn’t get rejected just because the trial uses a poorly selected cohort,” Alexander said.

The competition, called The Alzheimer's Disease Prediction of Longitudinal Evolution (TADPOLE) challenge, counts as its co-sponsors the Alzheimer's Association, Alzheimer's Society, and Alzheimer's Research U.K. Alexander said he is not sure yet how the five-figure prize money will be divvied up among categories—academic researchers versus high school students, he said, or teams who use imaging variables versus cognitive test scores. 

Researchers will test their prediction models on existing data that has been collected in ADNI1, ADNI-GO, and ADNI2 on cognitively normal people and others with mild cognitive impairment. These data come from cognitive tests, MRI, positron emission tomography of amyloid and glucose metabolism, and cerebrospinal fluid biomarkers. The contestants then use their models to predict which ADNI participants will progress clinically to Alzheimer’s dementia, decline on the ADAS-Cog13, and gain ventricular volume in MRI. The submission deadline is November 15, 2017. The study participants, who have agreed stay on for ADNI3, will be assessed on yet another visit through November 2018, and at that point TADPOLE will determine which winning algorithm best tracked their progression.

The first of three webinars for interested scientists was held July 12, and can be viewed here.—Gwyneth Dickey Zakaib