Data science in neurodegenerative disease: its capabilities, limitations, and perspectives

Area of Science:

• Computational neuroscience
• Biomedical data science
• Neurodegenerative disease modeling

Background:

• Advancements in computational approaches and biomedical data have spurred the development of diverse neurodegenerative disease models.
• Established models, while insightful, face limitations in clinical practice.
• Artificial intelligence (AI) presents a significant opportunity to address these deficiencies.

Purpose of the Study:

• To review the state-of-the-art in data-driven and knowledge-driven computational models for neurodegenerative diseases.
• To discuss the potential of AI in overcoming limitations of current models.
• To outline steps for integrating these models into clinical practice.

Main Methods:

• Review of diverse computational approaches including linear/nonlinear mixed models, differential equations, Cox-regression, Bayesian networks, and deep learning.
• Application of methods for understanding disease progression, predicting incidence, and stratifying patient subtypes.
• Integration of knowledge-based models (pathway-centric, knowledge maps) with data-driven analyses.

Main Results:

• Computational models provide granular insights into neurodegenerative disease progression and biomarker trajectories.
• AI-based approaches show promise in predicting disease incidence and identifying patient subtypes.
• Knowledge-driven models complement data-driven analyses by incorporating prior biological knowledge.

Conclusions:

• Computational models are relevant and useful tools in neurodegenerative disease research.
• AI has the potential to significantly improve the clinical utility of these models.
• Bridging the gap between current models and clinical application requires further development and validation.