Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clustering Alzheimer's Disease Subtypes via Similarity Learning and Graph Diffusion

Tianyi Wei, Shu Yang, Davoud Ataee Tarzanagh, Jingxuan Bao, Jia Xu, Patryk Orzechowski, Joost B. Wagenaar, Qi Long, Li Shen

TL;DR

Alzheimer's disease exhibits substantial heterogeneity that complicates diagnosis and treatment. The authors develop a framework combining SIMLR, a multi-kernel similarity learning method, with graph diffusion to jointly learn subject similarities and identify subtypes from MRI-derived cortical thickness features. Applied to data from the ADNI/TADPOLE cohort, they identify five AD/MCI subtypes with distinct biomarker and cognitive profiles, supported by targeted genetic associations (e.g., INPP5D), suggesting subtype-specific etiologies. The approach outperforms traditional clustering methods, driven largely by diffusion-based noise reduction, and is made available via public code, highlighting its potential for refining AD subtyping and guiding personalized interventions.

Abstract

Alzheimer's disease (AD) is a complex neurodegenerative disorder that affects millions of people worldwide. Due to the heterogeneous nature of AD, its diagnosis and treatment pose critical challenges. Consequently, there is a growing research interest in identifying homogeneous AD subtypes that can assist in addressing these challenges in recent years. In this study, we aim to identify subtypes of AD that represent distinctive clinical features and underlying pathology by utilizing unsupervised clustering with graph diffusion and similarity learning. We adopted SIMLR, a multi-kernel similarity learning framework, and graph diffusion to perform clustering on a group of 829 patients with AD and mild cognitive impairment (MCI, a prodromal stage of AD) based on their cortical thickness measurements extracted from magnetic resonance imaging (MRI) scans. Although the clustering approach we utilized has not been explored for the task of AD subtyping before, it demonstrated significantly better performance than several commonly used clustering methods. Specifically, we showed the power of graph diffusion in reducing the effects of noise in the subtype detection. Our results revealed five subtypes that differed remarkably in their biomarkers, cognitive status, and some other clinical features. To evaluate the resultant subtypes further, a genetic association study was carried out and successfully identified potential genetic underpinnings of different AD subtypes. Our source code is available at: https://github.com/PennShenLab/AD-SIMLR.

Clustering Alzheimer's Disease Subtypes via Similarity Learning and Graph Diffusion

TL;DR

Alzheimer's disease exhibits substantial heterogeneity that complicates diagnosis and treatment. The authors develop a framework combining SIMLR, a multi-kernel similarity learning method, with graph diffusion to jointly learn subject similarities and identify subtypes from MRI-derived cortical thickness features. Applied to data from the ADNI/TADPOLE cohort, they identify five AD/MCI subtypes with distinct biomarker and cognitive profiles, supported by targeted genetic associations (e.g., INPP5D), suggesting subtype-specific etiologies. The approach outperforms traditional clustering methods, driven largely by diffusion-based noise reduction, and is made available via public code, highlighting its potential for refining AD subtyping and guiding personalized interventions.

Abstract

Alzheimer's disease (AD) is a complex neurodegenerative disorder that affects millions of people worldwide. Due to the heterogeneous nature of AD, its diagnosis and treatment pose critical challenges. Consequently, there is a growing research interest in identifying homogeneous AD subtypes that can assist in addressing these challenges in recent years. In this study, we aim to identify subtypes of AD that represent distinctive clinical features and underlying pathology by utilizing unsupervised clustering with graph diffusion and similarity learning. We adopted SIMLR, a multi-kernel similarity learning framework, and graph diffusion to perform clustering on a group of 829 patients with AD and mild cognitive impairment (MCI, a prodromal stage of AD) based on their cortical thickness measurements extracted from magnetic resonance imaging (MRI) scans. Although the clustering approach we utilized has not been explored for the task of AD subtyping before, it demonstrated significantly better performance than several commonly used clustering methods. Specifically, we showed the power of graph diffusion in reducing the effects of noise in the subtype detection. Our results revealed five subtypes that differed remarkably in their biomarkers, cognitive status, and some other clinical features. To evaluate the resultant subtypes further, a genetic association study was carried out and successfully identified potential genetic underpinnings of different AD subtypes. Our source code is available at: https://github.com/PennShenLab/AD-SIMLR.
Paper Structure (19 sections, 4 equations, 6 figures, 6 tables)

This paper contains 19 sections, 4 equations, 6 figures, 6 tables.

Table of Contents

  1. INTRODUCTION
  2. MATERIALS AND METHODS
  3. Data Preparation
  4. Clustering and Graph Diffusion Methods
  5. Spectral Clustering.
  6. SIMLR.
  7. Graph Similarity Enhancement by Diffusion.
  8. RESULTS
  9. Cohort Description
  10. AD Subtypes Identification
  11. Selection of Cluster Number.
  12. Evaluation of clustering performances.
  13. Characterization of the clustered subtypes.
  14. Genetic Association Analysis of Identified AD Subtypes
  15. DISCUSSION
  16. ...and 4 more sections

Figures (6)

  • Figure 1: Schematic overview of the proposed methods (spectral clustering, SIMLR, and their diffusion variants) for AD subtype detection. For SIMLR+graph diffusion, one needs to repeat the similarity learning and graph diffusion steps for a few iterations.
  • Figure 2: Minimizing the separation cost gives the best choice of $K=5$ on the Laplacian of our data.
  • Figure 3: Similarity matrices for spectral clustering (SC) and SIMLR
  • Figure 4: Two-dimensional visualization with T-SNE for $K$-Means, SC, and SIMLR
  • Figure 5: Violin plots of biomarkers, demographical features, and cognitive test scores for each cluster and the control group (CN)
  • ...and 1 more figures