AlphaFold and AI Take On Alzheimer's After 20 Years Stuck on a Single Theory

Alzheimer's and Parkinson's remain among the most intractable diseases in medicine, despite trillions of dollars invested and decades of research. The causes are more complicated than they seem: biology is not the only obstacle to a cure. Science itself got in the way.

A diagnosis of peculiar cruelty

Terry Pratchett died in 2015 of posterior cortical atrophy, a rare form of early-onset Alzheimer's that attacks not memory but the visual cortex. He could see a cup on the table, but his brain could not process the signal. A writer who built entire universes out of images and words was struck by a disease that took away exactly that.

"I'm climbing forward with the others as hard as I can, to get there before the cure arrives," said

Pratchett in one of his last public interviews.

He donated money to research, made documentaries and called the disease "embuggerance," a damned inconvenient circumstance. He did not make it in time. And he was not the only one.

Twenty years stuck on a single theory

Since the early 1990s, Alzheimer's research has concentrated on the "amyloid hypothesis": the main cause of neuron death is plaques of the beta-amyloid protein. The idea proved fruitful for careers and grants, but not for patients. An informal power structure grew around it, known as the "amyloid mafia": research groups and editorial boards that promoted a single paradigm and pushed out alternatives such as the role of tau protein, neuroinflammation and mitochondrial dysfunction.

In 2022 a scandal broke. Key images in one of the most cited papers, a study by Sylvain Lesné, may have been falsified. The paper, on which hundreds of subsequent studies relied, came under investigation by the journal Science. The hypothesis did not collapse, but the monopoly began to crack.

The tally after three decades: more than a hundred failed clinical trials, billions of dollars written off, and a disease that still can be neither stopped nor reversed.

How AI is changing the rules

New tools attack the problem from several sides at once:

• Protein structures: Google DeepMind's AlphaFold has predicted three-dimensional models of millions of proteins, including those linked to neurodegeneration. Researchers are finding targets that could not previously be visualized.
• Early diagnosis: algorithms analyze speech patterns, eye movements and MRI data years before the first symptoms. Several systems exceed 80% accuracy.
• Molecule discovery: AI models generate and test millions of drug candidates in days instead of years.
• Failure analysis: AI finds subgroups of patients who responded to treatment in the data from failed trials, and proposes new hypotheses instead of burying the results.

What this means

AI has not found a cure for Alzheimer's. But for the first time in twenty years there are tools that search for one the right way: without the monopoly of a single theory, and with access to data that the human brain physically cannot process alone.

For the millions of people who are now where Pratchett was, this may be the most important news of the decade.