Researchers Identify Biological Differences Among Cognitively Defined Subgroups of Alzheimer Disease

December 14, 2018

The study found 33 additional single nucleotide polymorphisms with stronger relationships with a single subgroup.

Shubhabrata Mukherjee, PhD

A recent study published in Molecular Psychiatry has allowed 19 researchers from several institutions to create an approach to classify patients with Alzheimer disease in 6 subgroups, which is important for personalized medicine.

Researchers placed 4050 with late-onset Alzheimer disease into 6 groups based on cognitive functioning at the time of diagnosis, then used genetic data to uncover biological differences across these groups.

“This is in the realm of personalized medicine,” lead author Shubhabrata Mukherjee, PhD, research assistant professor, the University of Washington School of Medicine, told NeurologyLive. “Our study shows that Alzheimer is not one homogenous disease. So, a single drug might not be that effective for the whole population. We found a way to subtype Alzheimer based on their [patients] cognitive data at the time of diagnosis. Genome-wide genetic data enabled us to determine that a cognitively-defined categorization scheme produced biologically coherent subgroups of people with Alzheimer’s disease. This is an important result on the road towards personalized medicine.”

Researchers evaluated data from 5 studies where 2431 individuals had genome-wide single- nucleotide polymorphism (SNP) data. The mean age was 80; 92% self-reported white race, and 61% of participants were female. Study participants cognitive scores in 4 domains: memory, executive functioning, language and visuospatial functioning. Researchers then compared genotype frequencies for each subgroup to those from cognitively normal controls.

The largest group (39%) had scores in all 4 domains that were relatively close, while the next largest group (27%) had memory scores significantly lower than the other scores. Three of the smaller groups were identified by low language scores (13%), low visuospatial functioning scores (12%), and low executive functioning scores (3%). The final group (6%) had substantially lower scores in 2 of the 4 domains.

“There are a couple of implications here,” corresponding author Paul Crane, MD, professor of general internal medicine, University of Washington School of Medicine, explained to NeurologyLive. “Drug trials of Alzheimer’s drugs should consider whether the drug may work for 1 of the 6 subgroups. There may be a good response in 1 group that statistically would be drowned out by null responses in the other 5 groups. That would only be apparent if 1 looked separately at each subgroup. Of course, sample size may be an issue.”

Researchers also analyzed the subgroups to identify genetic variations that might explain the scoring patterns. Across the subgroups, there were 33 novel suggestive loci found across the genome with P < 10-5 and an extreme odds ratio in comparison to controls where none had statistical evidence of heterogeneity and 30 had odds ratios in the same direction across all datasets. These data back the biological coherence of cognitively defined subgroups, nominating novel genetic loci.

The second implication Crane explained is the identification of 33 new genetic signals associated with 1 of the subtypes. The genetic relationships were stronger than the strongest effects found by an earlier and larger international study where Alzheimer was treated as a single homogenous condition. “These provide targets for additional research that just might result in development of a treatment. Almost all of those targets had never previously been associated with any flavor of Alzheimer disease. For comparison, a huge international cohort studied 74,000 people (including all of the people in our paper) and found 20 genetic signals, each of which was weaker than all of the 33 genetic signals we found. We don’t know for sure, but we hope that labs following up on these signals — just as there are many labs around the world following up on the 20 undifferentiated Alzheimer disease signals from the earlier international paper – will identify several or many that are fruitful targets that will change the face of Alzheimer disease for people in that subgroup.”

The study also found a strong relationship between a particular variant of the APOE gene and risk for the memory loss subgroup.

The findings suggest the need to further evaluate relationships between cognitively defined late onset Alzheimer disease subgroups and rare Alzheimer subtypes.

“This work offers hope for a new path forward,” Crane concluded. “Despite the strong genetic data, this could prove to be yet another dead end in Alzheimer research; that is always a possibility, and given the fits and starts in this field over many decades, it is a very strong possibility, BUT — it’s also possible that this approach may be a critically important breakthrough that opens up a rapid stream of new discoveries and developments. We are hopeful that is the case! And it will be very exciting to see what happens next.”

Mukherjee explained that future plans include refining the subgroups as well as exploring how imaging correlates. While this sample specifically looked at European descent, study authors intend to incorporate additional studies from diverse ethnicities. Crane added that it’s important to explore the relationships between subtypes of typical Alzheimer disease and other rare Alzheimer disease variants like posterior cortical atrophy and logopenic primary progressive aphasia, as well as genetics of cognitively defined subtypes among those with early onset Alzheimer.


Mukherjee S, Mez J, Trittschuh E, et al. Molecular Psychiatry. Genetic data and cognitively defined late-onset Alzheimer’s disease subgroups. 2018.


: 1038/s41380-018-0298-8.