Research Paper: Nelvagal HR, et al. Brain: A Journal of Neurology, 2025. DOI: 10.1093/brain/awag114 | PMID: 41889331

Brains are weird, messy, and stubbornly refuse to fail in predictable ways. That's the problem researchers face when trying to understand why some people with Parkinson's disease develop dementia while others don't - and a new study just cracked open that mystery with machine learning, 399 post-mortem brains, and a whole lot of protein counting.

The Dementia Puzzle Nobody Could Solve

Lewy body diseases (LBD) - which include Parkinson's disease, Parkinson's disease dementia, and dementia with Lewy bodies - affect millions of people worldwide. Here's what makes them particularly frustrating to study: not everyone follows the same script. Some people accumulate toxic proteins and stay sharp. Others develop dementia with seemingly less pathology. Why?

The usual suspects have always been genetics (particularly the APOE gene), those clumpy alpha-synuclein proteins that define Lewy bodies, and Alzheimer's-like pathology that sometimes shows up uninvited. But figuring out how these factors interact? That's been like trying to solve a Rubik's cube in the dark.

Machine Learning Meets Microscopy

The research team, led by Hemanth Nelvagal and colleagues at University College London, decided to stop eyeballing brain slides and let algorithms do the heavy lifting. They built an image analysis pipeline using machine learning to automatically quantify alpha-synuclein, amyloid-beta, and phosphorylated tau across multiple brain regions.

The results were telling: their quantitative measurements correlated strongly with traditional semi-quantitative ratings but actually outperformed conventional staging methods in predicting who would develop dementia. Turns out, when you count proteins precisely instead of just rating them as "a lot" or "not much," you get better answers. Who knew.

The APOE Plot Twist

The APOE gene comes in different flavors - epsilon 2, 3, and 4. The epsilon 4 variant has long been known as the villain of Alzheimer's risk, and it plays dirty in Lewy body diseases too. But the relationship is more nuanced than "ε4 = bad."

Here's where it gets interesting: APOE ε3 carriers developed dementia at lower protein thresholds than ε4 carriers. Wait, what? The "protective" variant triggers dementia with less pathology?

Not exactly. The ε4 carriers still had higher overall dementia risk because they accumulated more pathological proteins in the first place. Think of it this way: ε4 is like having a leaky roof that floods your house, while ε3 is like having a house that can't handle even a small leak. Different failure modes, both problematic.

Blood Pressure, Brain Damage, and Sex Differences

The researchers also investigated orthostatic hypotension - that dizzy feeling when you stand up too fast that's common in Parkinson's. This blood pressure drop can starve the brain of oxygen, and the team found evidence of ischemic (low-oxygen) damage in many brains.

But here's the catch: orthostatic hypotension only increased dementia risk in ε3 carriers who had low amounts of Lewy and Alzheimer's pathology. For people already loaded with toxic proteins, the blood pressure issue didn't make things noticeably worse - they were already in trouble. And male sex added another layer of risk for this subgroup, because apparently men can't catch a break.

Four Roads to Ruin

Using an algorithm called SuStaIn (Subtype and Stage Inference), the team identified four distinct trajectories of Lewy pathology progression:

1. Brainstem-first: The classic pattern, starting in the lower brain and climbing upward
2. Amygdala-plus-brainstem: Early emotional brain involvement alongside the traditional starting point
3. Cingulate-brainstem combo: A new pattern with early involvement of the cingulate cortex
4. Neocortex-first: Starting in the thinking parts of the brain before spreading downward

Two of these patterns were already recognized. Two were new. The finding that Lewy pathology can start in the cortex and work its way down challenges the traditional "bottom-up" model of disease progression that's dominated the field for decades.

Why This Actually Matters

This isn't just academic protein-counting. Understanding these distinct pathways could transform how clinical trials are designed. Right now, researchers often lump all Lewy body disease patients together, then scratch their heads when treatments work for some people and not others.

With this kind of stratification - knowing someone's APOE genotype, their likely disease trajectory, and their co-pathology profile - trials could target specific subgroups. The ε3 carrier with low pathology and orthostatic hypotension needs a different approach than the ε4 carrier whose brain is accumulating proteins like they're on sale.

The Bottom Line

Dementia in Lewy body diseases isn't one disease with one cause. It's multiple pathways converging on the same terrible outcome. By combining old-school neuropathology with new-school machine learning, this research team has mapped out those pathways with unprecedented clarity.

The brain may be weird and messy, but it's starting to make a little more sense.

References

1. Nelvagal HR, Chiraki N, Curless T, et al. Quantitative pathology and APOE genotype reveal dementia risk and progression in Lewy body disease. Brain: A Journal of Neurology. 2025. DOI: 10.1093/brain/awag114. PMID: 41889331

2. Braak H, Del Tredici K, Rüb U, et al. Staging of brain pathology related to sporadic Parkinson's disease. Neurobiology of Aging. 2003;24(2):197-211. DOI: 10.1016/S0197-4580(02)00065-9. PMID: 12498954

3. Young AL, Marinescu RV, Oxtoby NP, et al. Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference. Nature Communications. 2018;9:4273. DOI: 10.1038/s41467-018-05892-0. PMID: 30323170

4. Coughlin DG, Hurtig HI, Irwin DJ. Pathological Influences on Clinical Heterogeneity in Lewy Body Diseases. Movement Disorders. 2020;35(1):5-19. DOI: 10.1002/mds.27867. PMID: 31483535

5. Bellomo G, Paciotti S, Gatticchi L, Parnetti L. The vicious cycle between α-synuclein aggregation and autophagic-lysosomal dysfunction. Movement Disorders. 2020;35(1):34-44. DOI: 10.1002/mds.27895. PMID: 31680350

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.