Classification and deep-learning–based prediction of Alzheimer disease subtypes by using genomic data
Problem
Late-onset Alzheimer's disease is heterogeneous, and genetic risk factors for its subtypes were previously unidentified, limiting subtype-specific prediction.
Why it matters
LOAD is the most common neurodegenerative disease among elderly people, and symptoms vary among patients.
What the AI did
A deep-learning prediction model was built on Japanese GWAS data to classify LOAD patients into two distinct genetic subtypes, one linked to major LOAD risk genes and one to kidney-disorder genes.
Results
Two subtypes identified in 1947 patients/2192 controls, validated in 847 patients/2298 controls; model accuracy not stated in source.
Impact
Evidence supports subtype discovery and a blood-test link to kidney function, but predictive performance is not shown.
Record
| Venue | Translational Psychiatry |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Daichi Shigemizu, Shintaro Akiyama, Mutsumi Suganuma, Motoki Furutani, Akiko Yamakawa, Yukiko Nakano, Kouichi Ozaki, Shumpei Niida |
| Citations | 37 (OpenAlex, at import time) |
| References | 42 |
| Model or method | deep-learning model; specific architecture not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41398-023-02531-1 |
| Open access | gold |
| Keywords | Alzheimer disease, GWAS, deep learning, subtypes, genomics, prediction |
Importance rating
major — millions of lives, an industry or a field
Abstract states LOAD is the most common multifactorial neurodegenerative disease among elderly people; subtype genetics were previously unidentified, so discovery aids understanding of a major disease.