Bayesian inference and role of astrocytes in amyloid-beta dynamics with modelling of Alzheimer's disease using clinical data

Alzheimer's disease (AD) is a prominent, worldwide, age-related neurodegenerative disease that currently has no systemic treatment. Strong evidence suggests that permeable amyloid-beta peptide (Abeta) oligomers, astrogliosis and reactive astrocytosis cause neuronal damage in AD. A large amount of Abeta is secreted by astrocytes, which contributes to the total Abeta deposition in the brain. This suggests that astrocytes may also play a role in AD, leading to increased attention to their dynamics and associated mechanisms. Therefore, in the present study, we developed and evaluated novel stochastic models for Abeta growth using ADNI data to predict the effect of astrocytes on AD progression in a clinical trial. In the AD case, accurate prediction is required for a successful clinical treatment plan. Given that AD studies are observational in nature and involve routine patient visits, stochastic models provide a suitable framework for modelling AD. Using the approximate Bayesian computation (ABC) approach, the AD etiology may be modelled as a multi-state disease process. As a result, we use this approach to examine the weak and strong influence of astrocytes at multiple disease progression stages using ADNI data from the baseline to 2-year visits for AD patients whose ages ranged from 50 to 90 years. Based on ADNI data, we discovered that the strong astrocyte effect (i.e., a higher concentration of astrocytes as compared to Abeta) could help to lower or clear the growth of Abeta, which is a key to slowing down AD progression.