Disease Knowledge Transfer across Neurodegenerative Diseases.

Area of Science:

• Neuroscience
• Biomedical Informatics
• Computational Biology

Background:

• Neurodegenerative diseases often present with limited multimodal biomarker data, especially rare conditions.
• Existing methods struggle to infer comprehensive biomarker trajectories from sparse, unimodal datasets.
• Posterior Cortical Atrophy (PCA) is a rare neurodegenerative disease with limited available multimodal data.

Purpose of the Study:

• To introduce Disease Knowledge Transfer (DKT), a novel technique for transferring biomarker information between related neurodegenerative diseases.
• To enable the inference of robust multimodal biomarker trajectories in rare neurodegenerative diseases using limited unimodal data.
• To apply DKT for estimating multimodal biomarker trajectories in Posterior Cortical Atrophy (PCA).

Main Methods:

• DKT is a joint-disease generative model that exploits shared biomarker relationships across diseases.
• The model was trained on a combined dataset including a large multimodal typical Alzheimer's Disease (tAD) dataset and a smaller unimodal PCA dataset.
• Validation was performed using synthetic data and patient datasets (TADPOLE and PCA cohorts).

Main Results:

• DKT successfully estimated ground truth parameters in simulations.
• The method predicted unseen biomarkers in both the TADPOLE and PCA patient datasets.
• DKT demonstrated the ability to estimate plausible multimodal biomarker trajectories in PCA, a disease with only unimodal MRI data.

Conclusions:

• DKT offers a powerful approach for inferring multimodal biomarker trajectories in rare neurodegenerative diseases with limited data.
• The technique is generalizable to other related neurodegenerative diseases beyond Alzheimer's variants.
• DKT has the potential to significantly advance biomarker discovery and diagnostic capabilities in neurodegenerative research.