A statistical and biological approach for identifying misdiagnosis of incipient Alzheimer patients using gene...

Area of Science:

• Bioinformatics
• Genomics
• Neuroscience

Background:

• Accurate diagnosis of Alzheimer's disease (AD) is crucial for effective treatment and research.
• Gene expression data holds potential for improving diagnostic accuracy.
• Misclassification of subjects can reduce the statistical power of studies.

Purpose of the Study:

• To develop and validate a misclassification algorithm for identifying potentially misdiagnosed Alzheimer's disease (AD) subjects using gene expression data.
• To assess the impact of correcting misclassifications on diagnostic model performance and the identification of AD-related genes.

Main Methods:

• Implementation of a latent-threshold model and a misclassification algorithm.
• Analysis of gene expression data from 16 Alzheimer's disease (AD) subjects.
• Comparison of model predictive power before and after adjusting for identified misclassifications.
• Mixed model analysis to detect differentially expressed genes.

Main Results:

• The initial model showed limited predictive power without the misclassification algorithm.
• The algorithm identified four subjects as potentially misdiagnosed.
• Adjusting the classification of these four subjects significantly increased the model's predictive ability.
• Mixed model analysis identified multiple AD-related genes only after correcting classifications.

Conclusions:

• The developed algorithm effectively identifies potentially misclassified subjects in Alzheimer's disease (AD) studies.
• Correcting misclassifications enhances the predictive power of diagnostic models.
• This approach improves the ability to identify disease-related genes, advancing Alzheimer's disease (AD) research.