Mobile Menu

Modelling Alzheimer’s Disease Progression to Simulate Intervention Strategies

A recent study presented SimulAD, a novel computational model of Alzheimer’s disease progression, which simulates intervention strategies for treatment of the condition.

Alzheimer’s disease progression

The number of people affected by Alzheimer’s disease recently exceeded 46 million, a figure that is expected to double every 20 years. Moreover, the disease mechanisms remain largely unknown and there are still no effective pharmacological treatments leading to tangible improvements of patients’ clinical progression. One of the major challenges in understanding Alzheimer’s disease is that its progression goes through a silent asymptomatic phase that can last decades before a clinical diagnosis is reached. To help in the design of appropriate intervention strategies, hypothetical models of the disease history have been proposed. These models characterise the progression of the disease by a cascade of morphological and molecular changes that affect the brain, which lead to cognitive impairment.

SimulAD: Modelling Alzheimer’s disease progression

This study presented SimulAD, a novel computational model of Alzheimer’s disease progression that analyses the potential effects of amyloid modifiers on the progression of brain atrophy, glucose metabolism, and the clinical outcomes for different scenarios of intervention. The researchers modelled the joint spatio-temporal variation of different modalities along the history of Alzheimer’s disease by identifying a system of Ordinary Differential Equations (ODEs) governing pathological progression. The ODEs system is specified with an interpretable low dimensional space relating multi-model data, and combines clinically-inspired constraints with unknown interactions that the researchers wish to estimate. To ensure interpretability of the relationships in the latent space, the researchers mapped each data modality to a specific latent coordinate.

SimulAD using the Bayesian framework

The model uses a Bayesian framework, where the latent representation and dynamics are efficiently estimated through stochastic variational inference. To generate hypothetical scenarios of amyloid lowering interventions, the researchers applied the SimulAD model to multi-modal imaging and clinical data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI).

The model transforms baseline multi-modal imaging and clinical information for a given subject into a latent variable composed of four z-scores. These z-scores respectively quantify the overall severity of atrophy, glucose metabolism, amyloid burden, and cognitive and functional assessment. The model then estimates the dynamical relationships across these z-scores to optimally describe the temporal transitions between follow-up observations. The transition rules are then mathematically defined by the parameters of Ordinary Differential Equations (ODEs), which are estimated from the data. This system allows researchers to compute the evolution of the z-scores over time from a baseline observation, and to predict the associated multi-modal imaging and clinical measures. SimulAD therefore enables researchers to simulate the pathological progression of biomarkers across the entire history of the disease. Moreover, now that the model is estimated, the researchers can modify the ODEs parameters to simulate different evolutionary scenarios according to different hypotheses.


This study presented the SimulAD model a novel quantitative instrument for the development of intervention strategies for disease modifying drugs in Alzheimer’s disease. The team’s framework enables the simulation of the effects of invention time and drug dosage on the evolution of imaging and clinical biomarkers in clinical trials. Moreover, the researchers state that SimulAD could be extended to account for panels of risk factors, which would ultimately allow clinicians to test in silico personalised intervention strategies.

Image credit: kjpargeter – FreePik

Share this article