Browse all




Big data framework seeks treatment targets for Alzheimer’s disease

13 Jun 2018 Samuel Vennin 
Research team

There is currently no cure for Alzheimer’s disease (AD), a progressive neurodegenerative disorder that affects about 50 million people worldwide. Hallmarks of the disease such as toxic amyloid-beta (Aβ) plaque aggregation or abnormal tau protein tangles are well known. However, the molecular mechanisms underlying AD neuropathology remain unknown and it is critical to identify positive biomarkers. Researchers from the US have resorted to big data analysis to find potential targets for future AD treatments (bioRxiv 10.1101/302737).

AD and AI

A team from the University of Washington, led by Su-In Lee and Matt Kaeberlein, undertook the challenge of developing a probabilistic model-based framework for identifying robust expression markers. They called this framework DECODER (discovering concordant expression markers) and applied it to AD. The framework was built from a meta-analysis combining three different studies that used a total of nine brain regions to form a pool of data that the model could draw conclusions from.

In order to use the whole database, regardless of tissue origin, the researchers first had to establish common features in each brain region. Comparing the overlap between the top 1000 Aß-associated genes in each region allowed them to hypothesize that basic mechanisms leading to the development of the disease were common across regions.

Three scores were then generated to quantify gene-concordant associations with neuropathology levels (such as Aβ levels) in multiple brain regions. The researchers found that global concordance-based scores were statistically more robust and informative than scores computed from each individual area. The top-scoring genes were also more likely to be part of a 144-gene AD pathway taken as a reference, which highlighted the biological relevance of the designed scores.

Identification and validation

Repeating the same approach for other pathways related to neurodegenerative diseases revealed that of all the top-scoring genes tested, only NDUFA9 was common to all pathways. This gene is part of a Complex I subunit in the mitochondria and plays a big part in mitochondrial respiration and synthesis of adenosine triphosphate (ATP).

To confirm the role of NDUFA9 in AD, the team carried out experiments on an animal model. A transgenic worm line was engineered to develop AD-like pathologies in its muscle cells at a specific stage in its life cycle. This leads to observable paralysis of the worms. By feeding the worms with bacteria that delivered an interfering RNA (RNAi), the researchers were able to inhibit the expression of NDUFA9.

The results were highly encouraging –  RNAi feeding strongly reduced Aβ plaque toxicity and significantly delayed any paralysis. Further experiments showed that altering any of other 13 Complex I subunits also delayed paralysis and significantly suppressed Aβ toxicity. This puts further emphasis on the importance of mitochondrial function in the development of AD.

Could it work in humans?

The next big challenge will be to replicate these results in humans. There are some major differences between human mitochondria and those of the worms used in the study, which makes the extrapolation of these results in worms to humans far from straightforward. More importantly, while partial inhibition of Complex I might be protective, Complex I also plays an integral part in the correct functioning of mitochondria. A balance will have to be achieved.

This study introduces a framework that will only get more powerful with time. As more AD studies on brain gene expression and neuropathology are published, the learning sample size of the framework will increase. The framework also has the potential to be applied to other pathologies such as cancers.


Copyright © 2018 by IOP Publishing Ltd and individual contributors
bright-rec iop pub iop-science physcis connect