Alzheimer’s disease is known to affect people along a spectrum. Some people get Alzheimer’s and rapidly deteriorate, others have Alzheimer’s for years without many difficulties. The underlying physiology of Alzheimer’s is extremely diverse, and researchers have posited that this could be a reason for the limited success of Alzheimer’s drug trials.
Using cerebrospinal fluid analysis data, a research team from the Netherlands led by Betty Tijms and Pieter Jelle Visser were able to successfully identify five separate molecular subtypes of Alzheimer’s disease. They analysed samples from 419 Alzheimer’s patients and compared those results to 187 healthy people, as a control.
The subtypes had distinct molecular characteristics – for instance, one subtype had increased production of amyloid plaques, whereas another subtype had reduced production of amyloid plaques but disruption of the blood-brain barrier. These differences at the molecular level were also reflected in the progression rate of the disease experienced by patients, and length of average survival time. The researchers also demonstrated that each subtype had a different genetic risk profile. The differences observed could mean that subtypes could respond differently to various treatments.
“There’s more than 140 drug targets that are being studied for Alzheimer’s disease and we can see indications in our data that a particular target might only work for a subgroup of patients,” says Tijms. “Clinical trials are testing the same drug on all of the Alzheimer’s patients without considering those subgroups and so if it only works for one subgroup and not for the others, then it is more difficult to see treatment effects. Now we know the subgroups exist, we can test for this.”
The group then validated their findings by creating a “subtype detector” using machine learning, and running it through a series of 6 cohort datasets from Europe and the US, where it identified subtypes with high accuracy.
Tijms and Visser are both working on the IMI project European Platform for Neurodegenerative Diseases (EPND). The individuals involved in the study were selected from a number of cohorts in the EPND Catalogue, and the data generated will be linked to EPND to enrich the platform and allow for further sharing and linking to other data. The researchers say that one of the next steps will be to try to identify subtypes not only in Alzheimer’s but also in other neurological diseases, such as Parkinson’s and dementia with Lewy bodies, where data are much sparser. The EPND catalogue will be extremely useful for furthering these studies.
“It’s not easy to collect cerebrospinal fluid samples from so many people, all with different diseases,” says Tijms, who is the lead author of the paper. “You need very large, high-quality patient samples – such as those found in the EPND catalogue – and by adding all of this data together we might be able to find some cross-disease mechanisms which could be shared. The more data the better!”
“If you look at the cohorts that are affiliated to EPND, we have around 20,000 samples. Ideally we would use them all, but time and money is limited. However, that illustrates the opportunities that EPND can offer,” says Visser.
The next steps will be to find similar subtypes in Parkinson’s and dementia with Lewy bodies. “We’re going to repeat this discovery exercise,” says Tijms. “Now we don’t know whether there are subtypes in Parkinson’s and we don’t have any tools to identify them. This is where EPND is at the frontier - we are able to find those groups that are large enough to start this subtyping."
EPND is supported by the Innovative Medicines Initiative, a partnership between the European Union and the European pharmaceutical industry.