EN / DA
Disease and treatment

Computer models can predict when high-risk individuals will develop Alzheimer’s

Computer models can become effective tools for making prognoses for people genetically predisposed for Alzheimer’s disease. In the future, doctors will be able to tell such people when they will develop this debilitating and deadly disease. The research can also be used to assess how effective treatments are.

Just mentioning Alzheimer’s to someone older than 60 years can send shivers down their spine, and for good reason, because more than 30 million people worldwide have this disease.

For most people, Alzheimer’s occurs for unknown reasons, which are probably somewhat associated with the environment and lifestyle. However, about 4% of the people who develop Alzheimer’s have a familial form of the disease related to genetics.

New Danish research shows that computer models can help predict whether this group of people who are genetically predisposed for Alzheimer’s will develop it at 70 years, 30 years or somewhere in between.

“Researchers have identified more than 200 mutations that are associated with an increased risk of familial Alzheimer’s. We analysed each of these mutations to determine whether we can predict at what stage in a person’s life each of these mutations will be expressed in the form of Alzheimer’s,” explains Kasper Planeta Kepp, Professor, Department of Chemistry, Technical University of Denmark, Lyngby.

The research results have been published in the Journal of Chemical Information and Modeling.

Mutations change plaque formation in the brain

The more than 200 Alzheimer’s-related mutations are located in the gene that encodes for the protein presenilin 1.

Presenilin 1 is a subcomponent of the gamma-secretase protein complex of the cell membrane, which causes the brain cells to produce amyloid-beta.

The people with Alzheimer’s form plaque with amyloid-beta 42 rather than amyloid-beta 40, and the Alzheimer’s-related mutations in the gene for presenilin 1 cause the difference between these two forms of amyloid-beta.

“However, how the various mutations alter the function of presenilin 1 and how this determines the severity of Alzheimer’s and the time of life when onset occurs were not previously known,” explains Kasper Planeta Kepp.

Tested 19 computer models

The researchers from the Technical University of Denmark trawled through genetic databases to collect data on each of the mutations that is associated with the onset of Alzheimer’s. The genetic data reveal which amino acids comprise presenilin 1 and thus also which should not be there or are in the wrong position in the protein.

Not all mutations in presenilin 1 cause Alzheimer’s. Some mutations do not noticeably alter its function. However, the more than 200 mutations investigated in this study did.

The researchers linked the genetic data with clinical data describing the pathway for people with each of these mutations, including the age at which they were diagnosed with Alzheimer’s.

The researchers then used the average clinical age of symptom onset for everyone who developed Alzheimer’s who had each mutation to test the computer models.

The researchers tested 19 computer models that can calculate how each mutation affects the function of presenilin 1.

Some mutations can have major effects if an altered amino acid destroys the stability of a region of presenilin 1 and thus also affects the function of the entire protein. Other mutations are less important if an amino acid is replaced by another one that affects the protein similarly.

“In addition, the position of the mutation is important. If the mutation is in an active part of the protein that is linked with the cleaving of amyloid-beta, this may explain why the cleavage is in the wrong location, resulting in amyloid-beta 42 rather than amyloid-beta 40,” says Kasper Planeta Kepp.

Some models excel in making prognoses for the various mutations

The researchers evaluated the ability of the various methods to precisely calculate the outcomes for each mutation and found that many were not very good at predicting the risk of developing Alzheimer’s.

The models were poor at estimating the actual progression of disease and the time of diagnosis.

But other methods excelled at predicting the effect of having a specific mutation.

Since the researchers had used all 19 computer models on all mutations, they could analyse the features of the models that succeeded in making prognoses. This gave the researchers insight into which molecular properties are expressed as the worst clinical outcomes.

Some changes in presenilin 1 are more important than others

The models that focus on the magnitude of the change to an individual amino acid caused by a mutation generally excelled in making prognoses.

The researchers therefore concluded that mutations that cause two amino acids differing greatly in chemical properties to be exchanged strongly affect the prognosis of Alzheimer’s.

Prediction improves even more when combined with mutations that cause structural changes in the protein.

For example, Kasper Planeta Kepp says that mutations involving proline and glycine instead of other amino acids almost always cause problems.

“There are many of these among the mutations that are linked to the development of familial Alzheimer’s. We also found that mutations in the alpha-helix protein structure cause problems. Our study has now enabled us to identify which parts of the computer models can enhance the value of the prognoses,” explains Kasper Planeta Kepp.

Using computer models clinically

According to Kasper Planeta Kepp, a few methods were so precise that they can be used in the future to estimate at what age people with these mutations will develop Alzheimer’s, especially considering differences between people that also contribute to the overall disease outcomes.

Kasper Planeta Kepp says that, although many factors affect the risk of developing Alzheimer’s at different stages of life, some mutations are so potent that other factors become less important.

Excellent computer models may make this type of analysis clinically important for people prone to familial Alzheimer’s and for physicians in the future. Early diagnosis and a strategy based on the severity of the disease for each individual are essential to optimize treatment strategies for Alzheimer’s.

People with a high-risk profile for early-onset Alzheimer’s may be advised to try to limit the influence of other risk factors on the progression of the disease.

“Alzheimer’s is caused by a combination of genetic and environmental factors. Although it is an unpleasant topic to discuss, early intervention in lifestyle and medicine may extend life by many years,” says Kasper Planeta Kepp.

Identifying targets for new medicines

Kasper Planeta Kepp says that excellent computer models may also become important tools in research attempting to understand Alzheimer’s.

Computer models may help to identify which mutations are most important for developing familial Alzheimer’s, and this knowledge can present new therapeutic targets.

“We need to understand how things go wrong when the disease progresses, and if we can see, for example, that many people develop Alzheimer’s because mutations occur in a specific protein structure, we can try to develop medicine that precisely targets the stability in that protein structure. Many large pharmaceutical companies have recognized that the production of amyloid-beta in the brain cells does not need to be limited, but the production of amyloid-beta 42 does. We still do not understand this completely, but computer models, such as the ones we have investigated, seem to indicate some, but not all, of the necessary elements that should be included in this type of medicine,” says Kasper Planeta Kepp.

Computing the pathogenicity of Alzheimer’s disease presenilin 1 mutations” has been published in the Journal of Chemical Information and Modeling. In 2017, the Novo Nordisk Foundation awarded a grant to Kasper Planeta Kepp for the project Predicting Clinical Outcome of Mutations in Alzheimer’s Disease.

Kasper Planeta Kepp
Professor
We are interested in understanding the natural evolution of proteins all the way from the individual amino acid changes, via the biophysical properties they give rise to, via the cell physiology, to the organism performance. We call this concept "From Sequence to Survival". Our first case study has been the evolution of heme and myoglobin. We now understand the molecular evolution that has tuned heme to facilitate reversible spin crossover binding of oxygen, and how the protein in myoglobin optimizes the binding of oxygen. We have also shown that oxygen storage and transport are too distinct functions manifesting differently in different mutants. Furthermore, while oxygen binding is similar in most mammalian myoglobins, its abundance differs by more than 10-fold. We have shown that this correlates with protein stability and that molecular evolution of stability of highly expressed myoglobin is a key event in whales as they strived to become deep divers.