Dynamic causal discovery in Alzheimer's disease through latent pseudotime modelling

2025-11-06

AI Summary

Most existing causal discovery methods rely on static graph assumptions, limiting their ability to characterize the dynamic pathological progression of neurodegenerative disorders such as Alzheimer’s disease (AD). To address this, we propose a dynamic causal discovery framework that integrates data-driven pseudotime inference with domain-knowledge-guided modeling. First, we employ a latent variable model to infer disease-specific pseudotemporal trajectories from plasma biomarkers (e.g., NfL, GFAP). Second, we construct time-varying causal graphs to explicitly model how causal relationships among biomarkers evolve across disease stages—thereby relaxing the static causal assumption. Our pseudotime metric achieves an AUC of 0.82 for AD diagnosis, substantially outperforming chronological age (AUC = 0.59). This work presents the first characterization of dynamic causal interactions between novel and conventional AD biomarkers during disease progression, yielding an interpretable, temporally grounded causal model that advances mechanistic understanding and supports early therapeutic intervention.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Studies of user behavior, including longitudinal effects of personalized systems
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

The application of causal discovery to diseases like Alzheimer's (AD) is limited by the static graph assumptions of most methods; such models cannot account for an evolving pathophysiology, modulated by a latent disease pseudotime. We propose to apply an existing latent variable model to real-world AD data, inferring a pseudotime that orders patients along a data-driven disease trajectory independent of chronological age, then learning how causal relationships evolve. Pseudotime outperformed age in predicting diagnosis (AUC 0.82 vs 0.59). Incorporating minimal, disease-agnostic background knowledge substantially improved graph accuracy and orientation. Our framework reveals dynamic interactions between novel (NfL, GFAP) and established AD markers, enabling practical causal discovery despite violated assumptions.

Problem

• Modeling evolving causal relationships in Alzheimer's disease progression
• Inferring disease trajectory using pseudotime independent of chronological age
• Discovering dynamic interactions between novel and established biomarkers

Innovation

• Latent pseudotime models disease progression dynamically
• Learns evolving causal relationships along data-driven trajectory
• Incorporates background knowledge to improve graph accuracy