Translatability of Animal Models for Alzheimer's Disease Using a Machine Learning Based Workflow.

Area of Science:

• Neuroscience
• Computational Biology
• Genomics

Background:

• Alzheimer's disease (AD) lacks effective disease-modifying therapies despite extensive research.
• Current animal models for AD may not accurately reflect human disease pathophysiology, hindering therapeutic development.
• Identifying translatable pathways is crucial for selecting relevant preclinical models.

Purpose of the Study:

• To develop and apply a machine learning workflow to assess the translational relevance of AD animal models.
• To identify shared dysregulated pathways between AD animal models and human datasets.
• To evaluate the predictive power of this workflow using a known therapeutic failure (ibuprofen).

Main Methods:

• Modified a machine learning workflow to identify translatable pathways across species.
• Analyzed hippocampal microarray data from three common AD mouse models (APP/PS1, 3×Tg, 5×FAD).
• Applied the workflow to publicly available data from ibuprofen-treated mice to predict clinical outcomes.

Main Results:

• No translatable pathways were identified in the APP/PS1 and 3×Tg models.
• The 5×FAD model exhibited translatable pathways, including SREBP control of lipid synthesis and cytotoxic T-lymphocyte pathways.
• The workflow accurately predicted the clinical failure of ibuprofen in human AD trials based on mouse data.

Conclusions:

• The 5×FAD model shows greater translational potential than APP/PS1 and 3×Tg models for Alzheimer's disease research.
• Evaluating the translatability of animal models is essential for successful therapeutic development in AD.
• The developed machine learning framework can improve the selection of preclinical models for Alzheimer's disease.