Benchmarking machine learning models for late-onset alzheimer's disease prediction from genomic data

Area of Science:

• Computational neuroscience
• Genetics
• Machine learning applications in medicine

Background:

• Late-Onset Alzheimer's Disease (LOAD) is a major cause of dementia with no cure.
• Preventive cognitive therapies are crucial, necessitating early risk estimation.
• Machine learning (ML) offers advanced tools for predicting LOAD risk.

Purpose of the Study:

• To systematically compare ML models for LOAD prediction.
• To utilize genetic variation data for risk assessment.
• To evaluate model performance using the ADNI cohort.

Main Methods:

• Comparison of representative ML models.
• Utilizing genetic variation data from the ADNI cohort.
• Performance evaluation using ROC curve analysis.

Main Results:

• The best ML models achieved approximately 72% area under the ROC curve.
• Demonstrated the efficacy of ML in predicting LOAD from genetic data.
• Identified top-performing models for LOAD risk estimation.

Conclusions:

• ML models show significant promise for estimating genetic risk in LOAD.
• Systematic model selection can aid in discovering novel genetic markers for LOAD.
• ML provides a valuable tool for advancing LOAD research and early intervention strategies.