General Discussion

LiberalArkie

(18,057 posts) Mon Apr 28, 2025, 08:26 AM Monday

AI Helps Unravel a Cause of Alzheimer's Disease and Identify a Therapeutic Candidate

April 25, 2025

A new study found that a gene recently recognized as a biomarker for Alzheimer’s disease is actually a cause of it, due to its previously unknown secondary function. Researchers at the University of California San Diego used artificial intelligence to help both unravel this mystery of Alzheimer’s disease and discover a potential treatment that obstructs the gene’s moonlighting role.

The research team published their results on April 23 in the journal Cell.

About one in nine people aged 65 and older has Alzheimer’s disease, the most common cause of dementia. While some particular genes, when mutated, can lead to Alzheimer’s, that connection only accounts for a small percentage of all Alzheimer’s patients. The vast majority of patients do not have a mutation in a known disease-causing gene; instead, they have “spontaneous” Alzheimer’s, and the causes for that are unclear.

Discovering those causes could ultimately improve medical care.

Snip

https://today.ucsd.edu/story/ai-helps-unravel-a-cause-of-alzheimers-disease-and-identify-a-therapeutic-candidate

3 replies

= new reply since forum marked as read

Highlight:

AI Helps Unravel a Cause of Alzheimer's Disease and Identify a Therapeutic Candidate (Original Post) LiberalArkie Monday OP

The AI they used is AlphaFold, which is (1) trained on a legal science data set, and (2) not generative AI highplainsdem Monday #1

There will probably be a multitude of good AI and datasets. I am waiting for the ones trained LiberalArkie Monday #2

Scan images are what the AI looks at. The "actual data" is a meaningless scramble Bernardo de La Paz Monday #3

highplainsdem

(55,681 posts)

1. The AI they used is AlphaFold, which is (1) trained on a legal science data set, and (2) not generative AI

Reply to LiberalArkie (Original post)

Mon Apr 28, 2025, 11:58 AM

Monday

like ChatGPT and other text generators, and image/video and music generators, which are illegally trained on all the intellectual property the AI companies could steal, and used (and intended) to compete directly with the creators whose work was stolen to train genAI.

This point needs to be made because the robber baron tech bros behind genAI like to fudge the distinction and pretend that if they aren't allowed to continue stealing all the writing, images, video and music they want to rip off, we won't have scientific advances.

LiberalArkie

(18,057 posts)

2. There will probably be a multitude of good AI and datasets. I am waiting for the ones trained

Reply to highplainsdem (Reply #1)

Mon Apr 28, 2025, 12:35 PM

Monday

on the data that CT and MRI scans generate and the diagnosis's that were found with the data generated.

Imagine going in for a scan and a report is generated with what was found from the actual data instead of what a radiologist guessed by looking at the imagines derived from the data.

Bernardo de La Paz

(55,127 posts)

3. Scan images are what the AI looks at. The "actual data" is a meaningless scramble

Reply to LiberalArkie (Reply #2)

Mon Apr 28, 2025, 12:51 PM

Monday

But your key point is good: there is much to be gained from AI finding things in image datasets.

It's the same reason that AI looks at pictures rather than the JPEG encoded or PNG encoded bytes.

The raw CT (Computerized Tomography) data is millions of data points from beams that are taken from all angles at all lengths. The beam data is correlated from the intersections of beams.

You could train an AI to understand the raw data or understand actual JPEG bytes, but it would be ridiculously inefficient because most of the training would be to replicate the construction of the images from the beam data or the JPEG bytes. The raw data is consumed by very mechanical, mathematically determined algorithms to form the images and that can be done 100% accurately in less time and with fewer computational units and with no training needed.

Reply to this discussion