Related Experiment Video
Updated: Oct 22, 2025
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.4K
Unsupervised Machine Learning to Identify Separable Clinical Alzheimer's Disease Sub-Populations
Jayant Prakash1,2, Velda Wang1, Robert E Quinn1,2
1Laboratory for Pathology Dynamics, Department of Biomedical Engineering, Georgia Institute of Technology and Emory University School of Medicine, Atlanta, GA 30332, USA.
Brain Sciences
|August 27, 2021
Summary
Unsupervised machine learning identified four Alzheimer's disease (AD) patient subgroups. These distinct clusters reveal patterns in cognitive performance, brain volume, and medication use, aiding future clinical trial design.
Area of Science:
- Neuroscience
- Computational Biology
- Biostatistics
Background:
- Alzheimer's disease (AD) patient heterogeneity complicates clinical trials and treatment evaluation.
- Identifying distinct AD patient subgroups is crucial for personalized medicine and research advancement.
Purpose of the Study:
- To define separable Alzheimer's disease (AD) clinical sub-populations using unsupervised machine learning.
- To characterize these sub-populations based on clinical features and medication usage.
Main Methods:
- Employed t-distributed Stochastic Neighbor Embedding (t-SNE) followed by k-means clustering on the ADNI dataset.
- Utilized Association Rule Mining (ARM) to identify patterns in patient demographics, neuroimaging, biomarkers, cognitive scores, and medications.
- Analyzed the ADNIMERGE dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI).
Main Results:
- Four distinct AD clinical sub-populations were identified based on cognitive performance and brain volume.
- Cluster-1: least severe disease; Cluster-0 & Cluster-3: mid-severity; Cluster-2: most severe disease.
- Association Rule Mining revealed medication patterns: anti-hyperlipidemia drugs with mid-severity AD, antioxidants with higher cognition, and antidepressants with severe AD. Vitamin D underutilization was noted across all groups.
Conclusions:
- The identification of four distinct AD sub-populations enhances patient stratification for clinical trials.
- Characterizing these subgroups by clinical features and medication associations provides valuable insights for targeted therapeutic strategies.
- This approach improves the precision of patient selection and comparative analysis across AD studies.
