AI models spot potential risk genes for Parkinson’s disease

The research from a team at the Cleveland Clinic Genome Center is detailed in npj Parkinson’s Disease.

AI was used to integrate and analyse multiple different forms of information from genetic, proteomic, pharmaceutical and patient datasets to identify patterns that may not be obvious from analysing one form of data on its own.  

“Parkinson’s disease is the second most common neurodegenerative disorder, right after dementia, but we don’t have a way to stop or slow its progression in the millions of people who live with this condition worldwide; the best we can currently accomplish is managing symptoms as they appear,” said study first author Lijun Dou, PhD. “There is an urgent need to develop new disease-modifying therapies for Parkinson's disease.” 

Making compounds that halt or reverse the progression of Parkinson's disease is especially challenging because the field is still identifying which genes cause which Parkinson’s disease symptoms when mutated, said Dr Dou.  

“Many of the known genetic mutations associated with Parkinson's disease are in non-coding regions of our DNA, and not in actual genes. We know that variants in noncoding regions can in turn impact the function of different genes, but we don’t know which genes are impacted in Parkinson’s disease,” she said. 

Using their integrative AI model, the team was able to cross-reference genetic variants associated with Parkinson’s disease with multiple brain-specific DNA and gene expression databases.

This allowed the team to infer which, if any, specific genes in the brain are affected by variants in non-coding regions of DNA. The team then combined the findings with protein and interactome datasets to determine which of the genes they identified affect other proteins in brains when mutated. They found several potential risk genes, such as SNCA and LRRK2, many of which are known to cause inflammation in the brain when dysregulated. 

MORE FROM ARTIFICIAL INTELLIGENCE

The research team next asked whether any available drugs could be repurposed to target the identified genes. Even after successful drugs are discovered and made, it can take an average of 15 years of testing for the medication to be approved.  

“Individuals currently living with Parkinson’s disease can’t afford to wait that long for new options as their conditions continue to progress,” said study lead and CCGC director Feixiong Cheng, PhD. “If we can use drugs that are already FDA-approved and repurpose them for Parkinson’s disease we can significantly reduce the amount of time until we can give patients more options.”  

By integrating their genetic findings with available pharmaceutical databases, the team found multiple candidate drugs. They then referenced electronic health records to see if there were any differences in Parkinson’s disease diagnoses for patients who take the identified drugs. For instance, individuals who had been prescribed the cholesterol-lowering drug simvastatin were less likely to receive Parkinson’s disease diagnoses in their lifetime.  

Dr Cheng said the next step is to test simvastatin’s potential to treat the disease in the lab, along with several immunosuppressive and anti-anxiety medications that warranted further study. 

“Using traditional methods, completing any of the steps we took to identify genes, proteins and drugs would be very resource- and time-intensive tasks,” said Dr Dou. “Our integrative network-based analyses allowed us to speed this process up significantly and identify multiple candidates which ups our chance of finding new solutions.”