Alzheimer’s: Can it be diagnosed shortly before the onset of symptoms? Researchers rely on AI

Drying point analysis
Can Alzheimer’s disease be diagnosed shortly before the onset of symptoms?

Until now, Alzheimer’s disease can only be diagnosed when there are symptoms. But then the neurons of the brain are already destroyed. The researchers are therefore betting on early detection with a simple method and artificial intelligence.

Researchers have developed a method that could be used in the future to diagnose Alzheimer’s disease at an early stage. Azam Jeihanipour and Jörg Lahann of the Karlsruhe Institute of Technology (KIT) can use it to determine the presence of pathologically altered proteins both in the blood and in the cerebrospinal fluid, i.e. in liquor samples, in a relatively simple way. Artificial intelligence helps them.

The differences are recognized thanks to artificial intelligence: the healthy beta-amyloid peptide (left) and a mutation of it (right).

(Picture: KIT)

But how does it work ? Misfolded beta-amyloid proteins are considered indicators of Alzheimer’s disease. It is assumed that even the smallest changes in the biochemical structure of proteins and peptides lead to the development of many neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease. The pathologically altered proteins can then no longer break down properly in the brain, clump together and ultimately damage neurons, leading to the well-known symptoms of the disease.

The research duo set out to find a method to identify the indicators. They came up with the idea of ​​detecting misfolding via the special drying structure of protein and peptide solutions. It is already known that the result of such drying spots on surfaces depends on the chemical properties of proteins. The speckle patterns that have been researched range from homogeneous films to branched and lattice patterns to complex arrangements. The method is also known as the coffee ring effect.

Pipetting robots and AI

In order to find out whether the method is also suitable for the detection of pathologically altered beta-amyloid proteins, the researchers dissolved diseased and healthy beta-amyloid proteins in a specific liquid and, for the required precision, dropped the solution on a pipette dripping window robot. Then the drops were dried for 40 minutes under controlled conditions.

However, since the differences in the drying images could hardly be distinguished by the human eye, Jeihanipour and Lahann decided to use artificial intelligence. The so-called deep learning system was initially “fed” with about 400 images of point patterns of the pathologically altered proteins and about 400 images of the healthy proteins. After that, the system was fed with 720 additional records for classification.

Spot patterns are like fingerprints

“The speckle patterns of beta-amyloid peptides represent fingerprints that reflect the structural and spatial identity of the peptide,” Lahann said in a KIT statement. They are not only characteristic and reproducible, but also lead to a classification of eight mutations with a prediction accuracy of more than 99%.

According to the researchers, this method has great potential. On the one hand, it could be used for further research on neurodegenerative diseases resulting from pathological proteins and peptides. The researchers also believe that the method could become a rapid, reliable and relatively simple diagnostic basis for identifying neurodegenerative diseases such as Alzheimer’s disease or Parkinson’s disease at an early stage. The results have been published in the current journal “Advances Materials”.

Leave a Comment