Predicting Alzheimer's Cognitive Resilience Score: A Comparative Study of Machine Learning Models Using RNA-seq Data.

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Alzheimer's disease (AD) is characterized by amyloid plaques and neurofibrillary tangles.
• Cognitive resilience (CR) describes preserved cognitive function despite AD pathology.
• Understanding CR mechanisms is crucial for developing effective interventions.

Purpose of the Study:

• To build and compare machine learning models for predicting CR scores.
• To identify genes and biological pathways associated with CR using RNA-seq data.
• To develop a tool for visualizing gene contributions to CR.

Main Methods:

• Utilized RNA-sequencing data from the ROSMAP and MSBB cohorts.
• Evaluated various machine learning models including SVR, random forest, XGBoost, linear, and transformer-based approaches.
• Employed Shapley additive explanations (SHAP) to identify important genes.

Main Results:

• The Support Vector Regression (SVR) model demonstrated superior performance in predicting CR scores.
• SHAP analysis identified specific genes contributing to cognitive resilience.
• Biological pathways linked to CR were elucidated through gene analysis.

Conclusions:

• Machine learning, particularly SVR, can effectively predict cognitive resilience in Alzheimer's disease.
• Gene expression patterns offer insights into the biological underpinnings of cognitive resilience.
• The developed Resilience Gene Analyzer (REGA) tool facilitates the interpretation of gene contributions to CR.