Subscribe to eTOC

Polygenic Hazard Score Predicts Amyloid Deposition and Alzheimer's Disease Neurodegeneration

Article In Brief

Investigators developed a system for predicting progression of Alzheimer's disease (AD)—a polygenic hazard score—based on data from several large population-based databases comprising genetic, imaging, clinical, and neuropathological data on healthy controls as well as patients with mild cognitive impairment (MCI) and those in the most severe stages of AD.

A team of scientists at University of California, San Francisco (USCF) and University of California, San Diego have developed a tool—a polygenic hazard score (PHS)—that uses genetic markers associated with Alzheimer's disease to determine whether the score can predict clinical and cognitive decline and brain pathology in older people.

The results suggest that the PHS is strong enough to identify disease signposts—before and after death—and the scientists believe that the test could ultimately help them assess how an individual's genetic fingerprint could be used to guide therapy. They are studying the tool for other neurologic disorders, as well.

“Like any new test, I think polygenic scores have to undergo rigorous evaluation before being adopted clinically,” said the senior study investigator, Rahul Desikan, MD, PhD, assistant professor of radiology & biomedical imaging, neurology and pediatrics at UCSF, and co-director of the Laboratory for Precision Neuroimaging, in an email to Neurology Today.

“Polygenic hazard scores must be tested for their ability to measure absolute risk for a single person from the general population. Provided with a collection of DNA samples, polygenic scores could retrospectively assess whether response to therapy is conditional on genetic risk in existing drug trials. Finally, polygenic scores should be examined for their ability to prospectively predict clinical outcomes (for example, the time to progression to dementia) in asymptomatic individuals. With PHS, we have shown that we can get at absolute risk and prospective prediction in older people who are cognitively normal.”

Of course, genes are only part of the story, he added. “Genetic susceptibility for complex conditions should not be viewed in isolation but as another risk factor for disease and combined with lifestyle and environmental factors in multivariate evaluation of disease risk. We are doing this right now and assessing how this score can be integrated with other markers.”

The findings from the study were published in January in Brain.

Study Design, Findings

To develop the PHS, Dr. Desikan, Chin Hong Tan, PhD, and their colleagues constructed a large database that houses anonymized genetic, imaging, clinical, epidemiology, and biomedical research data representing thousands of older people with no signs of cognitive problems as well as patients with mild cognitive impairment (MCI) and those in the most severe stages of AD. The collection also includes a large dataset of autopsy measurements.

The study group mined data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and other large longitudinal studies. ADNI has amassed clinical, imaging, genetic and biochemical markers in older adults with and without clinical disease to identify tools that would help detect the earliest signs of Alzheimer's and track progression over time. The UCSF scientists also used post-mortem data from the Religious Orders Study and Memory and Aging Project (ROSMAP) to see whether the polygenic score is associated with post-mortem amyloid plagues and tangles.

To generate PHS, they also analyzed genotyped data from 17,008 Alzheimer's disease patients and 37,154 controls from the International Genomics of Alzheimer's Project, and 6409 Alzheimer's disease patients and 9,386 older controls from the Alzheimer's Disease Genetics Consortium.

They identified 31 single nucleotide polymorphisms (SNPs) with the strongest ties to Alzheimer's. They used these SNPs to create their polygenic hazard score. They wanted to know whether a higher score, which is based on the number of SNPs an individual (or group) has, will predict cognitive decline and pathological load.

The first description and validation of the polygenic hazard score was published in PLOS Medicine in 2017. The researchers found that the score predicted AD dementia age of onset—even in people who don't carry the risk gene for AD, apolipoprotein E4 (APOE4) allele.

Last year, they reported in Acta Neuropathologica that positive amyloid PET scans and tau cerebrospinal levels varied as a function of the PHS. Those with a high PHS with amyloid and tau pathology had the sharpest longitudinal cognitive and clinical decline, even after controlling for the effects of APOE4. In the ROSMAP cohort, those with higher PHS had a greater amyloid load and more neurofibrillary tangles (at death). This occurred even in APOE4 noncarriers.

In the current Brain paper, they assessed whether PHS is associated with the biomarkers used to track AD, including regional amyloid PET, regional volume loss on MRI, post-mortem levels of regional amyloid plaque and tau tangles. They also used the score to see whether it was associated with other non-AD pathologies.


“Polygenic hazard scores must be tested for their ability to measure absolute risk for a single person from the general population. ...With PHS, we have shown that we can get at absolute risk and prospective prediction in older people who are cognitively normal.”


They computed a PHS on each individual, finding that the higher the PHS, the greater the rates of cognitive decline in the 632 non-demented patients in the analysis. There was a strong association between the score and the amyloid PET uptake volume ratio in frontal cortical regions in 980 AD patients and normal controls and with MRI volume loss in the entorhinal cortex in 607 AD patients and controls. They controlled for the strong genetic effect of APOE4.

The scores were also powerful predictors of the neuropathological findings. The research team evaluated autopsied brain tissue samples from 485 individuals who participated in the ROSMAP study and samples from 603 patients whose tissue is stored in the National Alzheimer's Coordinating Center.

The higher the PHS, the more amyloid load and tau tangles, the researchers observed. The PHS was also associated with Lewy body and neurovascular pathologies. The score was not associated with other non-AD pathologies they looked at, including medial temporal lobe sclerosis, and frontotemporal lobar degeneration (FTLD) pathology (Pick's disease, corticobasal degeneration, progressive supranuclear palsy, and FTLD with TDP-43).


“The strength of this investigation is that they are using the PHS to predict different endophenotypes (amyloid PET, MRI volume loss) and neuropathological measures, and validating their findings across different studies.”


PHS has utility for predicting cognitive and clinical decline even after accounting for the effects of imaging markers of amyloid accumulation and atrophy, the scientists said. The scientists believe that PHS could possibly be used to identify individuals at the highest risk for dementia.

“The polygenic component beyond APOE4 has utility for disease prediction,” said Dr. Tan, the first author of the study, who previously was a post-doc in Dr. Desikan's lab and remains an integral part of these studies. He is now on the faculty at Nanyang Technological University in Singapore.

Dr. Tan said that it is possible to stratify individuals based on their PHS into percentiles. “Genetics is a single piece of the puzzle and we are also interested in the role of other non-genetic factors in influencing AD progression.”

He said that PHS may be useful for reducing costs associated with high amyloid PET screening failure rates in preclinical AD trials. Researchers could use PHS to screen potential participants first and only conduct amyloid PET scans if they have a high PHS. PHS could also be used to enrich clinical trials by identifying biomarker positive individuals who are most at risk for short-term progression.

Expert Commentary

“There are a lot of different ways to do polygenic risk scores,” said David A. Bennett, MD, director of the Rush Alzheimer's Disease Center and the Robert C. Borwell professor of neurological sciences. He is a co-author of this study and other studies on polygenic risk scores because these collaborative studies require access to large databases.

“The field is working on prediction of clinical disease and prediction of pathological disease,” he said, but he cautioned: “Just because you have SNPs that are linked to the clinical phenotype doesn't mean they will predict AD pathology. In this case, it did.”

His team at Rush is now looking at motor phenotypes (parkinsonism and change in gait speed, for instance) that are also common in Alzheimer's.

Dr. Bennett does not think that these types of genetic scores alone “will guide clinical decision making,” but, he said, “they may help.”

“Complex disorders like late-onset Alzheimer's disease are associated with many risk variants that slightly increase susceptibility,” said Shea Andrews, PhD, a molecular geneticist and post-doctoral fellow in the Ronald M. Loeb Center for Alzheimer's Disease at the Icahn School of Medicine at Mount Sinai. “As such, aggregating the effect of many genes into a summary risk score for the disease can confer significantly more information in regard to an individual's risk of developing a disease in comparison to evaluating the effects of a single variant.”

“As individual genetic variants often have pleiotropic effects, it is a natural extension for the researchers to evaluate whether a PHS for clinically diagnosed AD is also associated with other phenotypes,” Dr. Andrews added. “By showing that a PHS for AD is not only associated with Alzheimer's disease but also the underlying pathological changes observed in dementia, it further emphasizes that polygenic risk scores can be used to identify individuals who are at risk of developing AD. Furthermore, by demonstrating associations with particular neuropathologies and not others, the results of this study can be used to infer the underlying biological pathways that genetic variants associated with AD may promote disease.”


PHS is associated with regional post-mortem neuropathology. Beta estimates of the associations between PHS and regional post-mortem (A) amyloid load and (B) neurofibrillary tangles.

Ultimately, he added, “polygenic risk scores can be used in conjunction with lifestyle and environmental risk factors to stratify individuals into high- and low-risk groups for inclusion into AD studies prior to the development of clinical symptoms and evaluate how lifestyle and pharmacotherapeutic interventions can alter the disease course.”

“The strength of this investigation is that they are using the PHS to predict different endophenotypes (amyloid PET, MRI volume loss) and neuropathological measures, and validating their findings across different studies,” said Giuseppe Tosto, MD, PhD, assistant professor of neurology at Columbia University Medical Center.

Dr. Tosto believes that it does have a potential application for personalized medicine. “You get an individual score and stratify a person into high-, low-, or medium- risk based on their genome.” He and his colleagues published a paper in 2017 in Neurology using PHS in two familial AD cohorts of non-Hispanic whites and Caribbean Hispanics.

Dr. Rahul Desikan Is on a Personal Mission

Around the same time that neuroradiologist Dr. Rahul Desikan began working on the PHS for Alzheimer's disease, he noticed he was developing upper and lower motor weakness, symptoms that were suggestive of a neurologic disorder.

In early 2017, his group identified two new SNPs that were strongly associated with amyotrophic lateral sclerosis (ALS). A few weeks later, he was diagnosed with ALS.

Dr. Desikan can no longer speak, walk, or use his hands. But he uses eye-tracking software to communicate with others, including his colleagues who now arrive at his house every day to continue their research.

Dr. Desikan feels a sense of urgency around these studies. He said he will continue his research for as long as he can.

Link Up for More Information

• Tan CH, Bonham LW, Fan CC, et al. Polygenic hazard score, amyloid deposition and Alzheimer's neurodegeneration Brain 2019; 142(2): 460–470.
• Desikan RS, Fan CC, Wang Y, et al. Genetic assessment of age-associated Alzheimer disease risk: Development and validation of a polygenic hazard score PLoS Med 2017; 14: e1002258.
• Tan CH, Desikan RS. Interpreting Alzheimer disease polygenic scores Ann Neurol 2018; 83(3): 443–445.
• Tan CH, Fan CC, Mormino EC, et al. Polygenic hazard score: an enrichment marker for Alzheimer's associated amyloid and tau deposition Acta Neuropathol 2018; 135(1): 85–93.
• Tan CH, Hyman BT, Tan JJX, et al. Polygenic hazard scores in preclinical Alzheimer disease Ann Neurol 2017; 82(3): 484–488.
• Tosto G, Bird TD, Tsuang D, et al. Polygenic risk scores in familial Alzheimer disease Neurology 2017;88(12):1180–1186.