By Heather D’Angelo and Frank Fuller
‍

Our newly-released Digital Twin Generator (DTG) for Alzheimer’s disease (AD) introduces some key enhancements, improving its ability to create precise digital twins of clinical trial participants. 

Expanded Patient Dataset and Comprehensiveness

Our patient training dataset has expanded from around 25,000 to 27,000 patients due to the inclusion of data from multiple sources, one of which includes the Critical Path for Alzheimer’s Disease (CPAD). This increase not only broadens our dataset but also ensures that our model benefits from recent clinical trial data that aligns with current research and treatment protocols, as acknowledged in our data sources. These trials often target early-stage Alzheimer's patients and incorporate new biomarkers and outcome measures. The enhanced model now offers longitudinal prediction of key biomarkers like amyloid and tau proteins, essential for understanding Alzheimer's progression. In addition, we now predict individual item scores for the Functional Activities Questionnaire, which is useful for differentiating disease progression in mild-severity patients. Overall, these expansions provide a more useful tool kit for mechanistic studies of Alzheimer’s Disease.

Improved Model Architecture

The update to a Pure Neural Boltzmann Machine marks a significant enhancement in our model’s architecture. This advancement boosts the model’s capability to model complicated patient trajectories, which overall has improved the accuracy of the outcome trajectories we predict. Enhanced modeling of these variations is critical for the development of personalized medicine, a growing focus in healthcare that aims to tailor treatments to individual patient needs more effectively.

We encourage you to download our AD-DTG 4.0 specification sheets to gain a comprehensive understanding of this model, including its capabilities and underlying data.