B-BIND: Biophysical Bayesian Inference for Neurodegenerative Dynamics.

Area of Science:

• Computational Biology and Neuroscience
• Biophysics
• Statistical Modeling

Background:

• Biological systems exhibit complex dynamics, with underlying states often unobservable directly.
• Inferring latent biological states is crucial for understanding diseases like Alzheimer's Disease (AD).
• Current AD research often relies on discrete snapshots of pathology, lacking continuous disease progression insights.

Purpose of the Study:

• To develop a biophysically motivated Bayesian framework (B-BIND) for continuous inference of neurodegenerative disease states.
• To model Alzheimer's Disease progression by inferring a pseudotemporal order of donors based on pathological burden.
• To analyze and refine the model by identifying the most informative pathological features.

Main Methods:

• Proposed a biophysically motivated Bayesian framework (B-BIND) for continuous inference of latent disease states.
• Modeled pathological burden as an exponential process and introduced pseudotime to order donors.
• Employed linearization for theoretical analysis of model convergence and identifiability, and Markov chain Monte Carlo for estimation.

Main Results:

• Demonstrated the effectiveness of the B-BIND framework through simulation studies under various data conditions.
• Applied the methodology to the Seattle Alzheimer's Disease Brain Cell Atlas, successfully inferring a pseudotemporal ordering of donors.
• Identified informative pathological features to refine the model for enhanced accuracy in neurodegenerative disease analysis.

Conclusions:

• The B-BIND framework provides a robust method for continuous pseudotime modeling in neurodegenerative disease research.
• This approach enables more accurate inference of disease states from discrete pathological observations.
• Lays the groundwork for advanced computational analysis of disease progression dynamics.