Key takeaways:
- Giant-data AI fashions can also be used to determine non-FDA-approved medication for Parkinson’s illness.
- There’s potential for AI fashions to determine Parkinson’s medication to deal with different non-neurodegenerative situations.
AI could present a solution to the query of whether or not medication initially developed and prescribed for non-neurodegenerative bodily situations can be utilized to deal with people with Parkinson’s illness.
There’s a urgent want to seek out efficient therapeutics for people with PD, as development of the illness results in diminished performance, whereas the medical trials that inform the regulatory course of for approval of disease-specific remedies can take greater than a decade to finish.
Feixiong Cheng, PhD, head of the Cheng Laboratory and director of the Cleveland Clinic Genome Middle, and his fellow researchers employed a technique referred to as techniques biology.
Utilizing their integrative AI mannequin, the researchers first cross-referenced PD-related genetic variants with a number of brain-specific DNA and gene expression databases. Subsequent, they mixed the findings with protein and interactome datasets to evaluate which of the recognized genes have an effect on proteins within the mind, yielding a number of recognized to trigger irritation when mutated.
The researchers then created an integrative AI mannequin that synthesized and analyzed a broad vary of knowledge from genetic, proteomic, pharmaceutical and affected person datasets to uncover patterns which will have been hidden if evaluation was undertaken on any one of many datasets individually. By combining these knowledge, Cheng and colleagues had been in a position to generate a listing of a number of candidate medication.
Subsequent cross-referencing of digital well being data revealed whether or not there have been variations in PD diagnoses for these prescribed the beforehand recognized medication, with evaluation discovering that people prescribed the FDA-approved, cholesterol-lowering drug simvastatin had been much less more likely to obtain PD diagnoses.
The analysis was revealed earlier this yr in NPJ Parkinson’s Illness.
Healio spoke with Cheng, who clarified the philosophy behind the method, whereas additionally offering a glimpse into the way forward for large-model, large-data AI utilization to find medication that may be repurposed not just for neurodegenerative situations, but additionally for a wider vary of medical situations.
Healio: What was the impetus behind enlisting AI/machine studying for this sort of process?
Cheng: We lack efficient remedies for PD, and there aren’t any good experimental fashions for PD drug discovery. AI-based computational strategies provide efficient methods for PD goal and drug discovery.
Healio: What gaps in conventional analysis does the help of AI shut or enhance for this objective?
Cheng: Conventional genome-wide affiliation research recognized many PD danger variants, however these danger variants are in non-coding human genomes, and we don’t know which genes or drug targets are linked to those non-coding PD danger variants. Through integration of useful genomics knowledge utilizing our AI fashions, we are able to determine seemingly causal genes for PD as potential drug targets for rising therapeutic growth.
Healio: Is there an general demand in well being take care of repurposing medication, and what made neurodegenerative situations a primary candidate for repurposing?
Cheng: Sure. There’s a excessive demand in well being take care of repurposing medication, together with ALS, Alzheimer’s illness, PD, and different illnesses. For instance, 33% of compounds below AD medical trials are repurposed medication primarily based on our 2025 AD pipeline.
Healio: Might the AI fashions be skilled to seek out non-FDA-approved medication that could also be used for a similar objective sooner or later?
Cheng: Sure. Past non-FDA medication, we’re additionally pursuing to determine medical failure medication (with good security profiles however missing results on one particular indication) or intestine metabolites generated in our human physique (just like the intestine as nicely) utilizing our superior AI know-how.
Healio: Do you foresee that the AI fashions might be skilled in reverse, ie, discovering out whether or not medication for situations reminiscent of PD could help therapy for points in different specialties?
Cheng: Sure. AI can determine many PD or AD brokers that failed in medical trials resulting from lack of efficacy however nonetheless had good security profiles for different illnesses reminiscent of ALS or different difficult illnesses as nicely.
Healio: For which different types of analysis may the AI fashions be utilized?
Cheng: Past drug repurposing, AI can be utilized to design novel small molecules or antibodies/nanobodies for some difficult persistent illnesses, reminiscent of AD, PD and most cancers. We’ve got a number of thrilling tasks to use generative AI to design novel antibodies/nanobodies as efficient organic therapeutics for AD as nicely.
References:
Synthetic intelligence mannequin identifies potential danger genes for Parkinson’s illness. https://www.lerner.ccf.org/information/article/?title=Synthetic+intelligence+mannequin+identifies+potential+danger+genes+for+ParkinsonpercentE2percent80percent99s+illness&id=834425e79fe1c1beb8f60901bb996ec3925df258. Revealed Jan. 27, 2025. Accessed June 2, 2025.
Cheng F, et al. Cell Rep. 2024;doi:10.1016/j.celrep.2024.114128.
Cheng F, et al. Nat Commun. 2025;doi:10.1038/s41467-025-56690-4.
Cheng F, et al. NPJ Parkinsons Dis. 2025;doi:10.1038/s41531-025-00870-y.
For extra data:
Feixiong Cheng, PhD, may be reached at neurology@healio.com.
