Random forest-integrated analysis in AD and LATE brain transcriptome-wide data to identify disease-specific gene...

Area of Science:

• Neuroscience
• Genetics
• Bioinformatics

Background:

• Alzheimer's disease (AD) and Limbic-predominant age-related TDP-43 encephalopathy (LATE) are common neurodegenerative disorders with overlapping symptoms.
• Understanding disease-associated genes is crucial for developing effective treatments, but imbalanced clinical data presents a significant analytical challenge.
• Existing machine learning algorithms struggle with imbalanced datasets common in neurodegenerative disease research.

Purpose of the Study:

• To develop a novel algorithm for identifying disease-associated genes from imbalanced transcriptome-wide data.
• To address the challenge of analyzing highly imbalanced clinical samples in neurodegenerative disease research.
• To facilitate the discovery of new biomarkers and therapeutic targets for AD and LATE.

Main Methods:

• Proposed an Integrated Multiple Random Forests (IMRF) algorithm designed for imbalanced transcriptome-wide data analysis.
• Utilized IMRF to differentiate genes associated with LATE and/or AD from control subjects.
• Validated the IMRF method through cross-domain verification, classification performance using identified genes, and independent testing.

Main Results:

• The IMRF algorithm effectively identified genes with altered expression in LATE and/or AD patients.
• Demonstrated competitive classification performance using the genes identified by IMRF.
• Confirmed the effectiveness of IMRF in handling imbalanced datasets for gene identification in neurodegenerative diseases.

Conclusions:

• IMRF is an effective feature selection algorithm for imbalanced data in neurodegenerative disease research.
• The identified genes hold promise for developing new gene biomarkers for AD and LATE.
• IMRF can facilitate the development of effective strategies for disease prevention and treatment.