Detecting Risk Gene and Pathogenic Brain Region in EMCI Using a Novel GERF Algorithm Based on Brain Imaging and...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Informatics

Background:

• Multi-modal data fusion is crucial for understanding complex brain diseases like early mild cognitive impairment (EMCI).
• Current machine learning methods struggle with high-dimensional, small-sample multi-modal data, limiting fusion feature selection accuracy.
• Identifying genetic risk factors and brain region associations is key to understanding EMCI pathogenesis.

Purpose of the Study:

• To develop a robust algorithm for fusion feature selection in multi-modal data for EMCI.
• To identify novel risk genes and disease-related brain regions associated with EMCI.
• To enhance the classification accuracy and reliability of machine learning models for complex brain diseases.

Main Methods:

• Proposed a genetic-evolutionary random forest (GERF) algorithm integrating genetic data and resting-state functional magnetic resonance imaging (rs-fMRI) data.
• Employed classical correlation analysis to construct fusion features between brain regions and genes.
• Introduced genetic-evolutionary principles to optimize feature extraction and improve classification performance.

Main Results:

• The GERF algorithm achieved satisfactory classification accuracy on the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, demonstrating effectiveness in small sample learning.
• Outperformed existing methods in multi-modal data analysis for EMCI.
• Successfully identified potential risk genes and brain regions associated with EMCI.

Conclusions:

• The proposed GERF algorithm offers a novel and effective approach for multi-modal data fusion and feature selection in EMCI research.
• This framework provides valuable insights for early diagnosis, clinicopathologic analysis, and potential therapeutic interventions for EMCI.
• The methodology can be extended to analyze multi-modal data for other complex diseases.