Researchers use AI to predict different drivers of Parkinson’s

The AI examined images of brain cells and was able to identify changes in key areas of the cell that are characteristic of the different known mechanisms that cause Parkinson’s.

How does Parkinson’s develop?

Parkinson’s is caused by the loss of dopamine producing cells in the brain. Although the exact cause is unknown, research has suggested that Parkinson’s is caused by a combination of genetics and the environment. But there are a lot of different pathways in the body that could be the key driver of Parkinson’s if they don’t work as they should.

2 of the pathways implicated in Parkinson’s are the build-up of toxic alpha-synuclein, and problems with energy-producing mitochondria, which means brain cells don’t get enough energy to survive.

The age at which people develop Parkinson’s, the symptoms that people experience, how fast symptoms progress and how severe people’s symptoms are, are all things that vary in each individual. Some researchers suggest that this is because Parkinson’s could be caused by problems in different pathways in the body. 

What is artificial intelligence?

Artificial intelligence (AI) is the intelligence of machines or software, instead of the intelligence of human beings or animals. One way it does this is by using data and algorithms to copy how humans learn. This improves the accuracy of the predictions it can make. It can be used for many different things, including to predict how certain conditions are caused.

Quick summary

In a new study, researchers examined images of different brain cells that were engineered to have defects in specific pathways, such as mitochondrial dysfunction. Read a summary of the research in Nature Machine Intelligence. Using AI they could accurately distinguish which pathways were defective in different cells. This opens up the possibility that AI could be used to subtype Parkinson’s in the future, to help develop more personalised treatments.

What did the researchers do?

Using stem cells (cells that can be made into any type of cells, such as heart cells, skin cells and brain cells), the researchers engineered brain cells which mimic 4 of the different pathways that are affected in Parkinson’s. The different pathways they mimicked were:

  • Brain cells with mutations in the SNCA gene, which causes alpha-synuclein to form toxic clumps.
  • Alpha-synuclein that travels from brain cell to brain cell, which increases how many cells are damaged by alpha-synuclein.
  • Brain cells with defective mitochondria, caused by a pesticide called rotenone.
  • Brain cells with mutations in the PINK1 and PARKIN genes, which cause problems with clearing away defective mitochondria. The PINK1 and PARKIN genes are associated with young onset Parkinson's.

The researchers took the brain cells and labelled the different parts of the cell using fluorescent markers. They labelled the mitochondria, the cell centre or nucleus and the lysosomes, which break down waste proteins from the cell.

They took images of the brain cells, which included the labelled parts of the cells. The AI looked at each of the images, as well as images of brain cells that were engineered without any problems. The AI looked at lots of different features in each of the brain cells and used this information to see how different parts of the cell interact with each other in cells with problems in different pathways, and in those which did not have problems.

After the AI understood how the cells worked, the researchers then showed the software images it hadn't seen before. This was to determine whether those cells had Parkinson’s, and if so, which pathway was altered in the cell to cause Parkinson’s.

What were the results?

The AI was able to determine whether the brain cell was normal or identify which of the 4 altered pathways in Parkinson’s it expressed with 95% accuracy.

The mitochondria and the lysosomes were the most helpful features that helped the AI detect the correct type of pathology. This suggests these changes in these 2 pathways may be key drivers of Parkinson’s.

What are the next steps?

These results are positive. But they are only from brain cells engineered from stem cells. The next steps will involve testing the AI on brain cells from people with Parkinson’s, to see if the results can be replicated.

The researchers will also look at other pathways that don’t work properly in people with Parkinson’s, and test if the AI can detect problems in more pathways accurately.

What do these results mean?

By understanding which pathways are not working correctly in people with Parkinson’s, in the future we may be able to use this information to work out if people have different subtypes of Parkinson’s. We could then use this to develop more personalised treatments and target different pathways in people with different subtypes of Parkinson’s, so treatments may be more effective.

In the future, drug therapies could be tested on brain cells engineered from stem cells to see which pathways they are best at treating. And if successful at treating a certain pathway that drives Parkinson’s, clinical trials could start looking at recruiting people depending on the pathways that are the key drivers of their Parkinson’s.

Professor David Dexter, Director of Research at Parkinson’s UK, said:

"This groundbreaking research has harnessed the potential of AI to identify unique features in brain cells grown in the laboratory, which are characteristic for different drivers of Parkinson’s, such as mitochondrial dysfunction. This raises the possibility that one day we can subtype Parkinson’s and develop personalised medicines to stop, slow or reverse symptoms.

"Whilst this is an important step, researchers still need to demonstrate whether this technology can reliably subtype people living with Parkinson’s. If that’s possible, its practical application in clinical trials could see more new life changing treatments be developed and come to market. Something the 145,000 people living with the condition in the UK want and deserve."

Join the Research Support Network

Sign up to receive emails with the latest research news, events and opportunities to take part in research.