Classification of genes and putative biomarker identification using distribution metrics on expression profiles

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Identifying genes with switch-like expression is crucial for understanding gene regulatory mechanisms.
• Such genes are likely associated with tissue-specific expression, making them potential biomarkers.

Purpose of the Study:

• To systematically classify genes based on their expression profiles.
• To identify candidate tissue-specific biomarker genes using a novel computational approach.

Main Methods:

• Analyzed over 16,000 gene expression profiles from 2,145 mouse array samples.
• Classified gene expression profiles into four categories: predominantly-off, predominantly-on, graded, and switch-like, using distribution metrics.
• Compared gene expression profiles between tissue groups (e.g., brain vs. non-brain) using Kolmogorov-Smirnov distance and Pearson correlation.

Main Results:

• Successfully classified genes based on their expression patterns.
• Identified potential tissue-specific biomarker candidate genes by analyzing differences in gene expression between tissues.
• Demonstrated a method for systematic gene classification and biomarker discovery.

Conclusions:

• The developed methodology enables the identification of switch-like genes.
• This approach facilitates the discovery of tissue-specific biomarkers.
• The methodology holds potential for advancing disease-specific biomarker discovery.