A problem with the correlation coefficient as a measure of gene expression divergence

Area of Science:

• Genomics
• Bioinformatics
• Comparative Genomics

Background:

• Gene expression profiles are crucial for understanding species divergence.
• The correlation coefficient is a common metric for assessing gene expression divergence.
• Potential biases in commonly used statistical methods can impact biological interpretations.

Purpose of the Study:

• To identify limitations of the correlation coefficient in measuring gene expression divergence.
• To evaluate alternative metrics for assessing expression divergence in uniformly expressed genes.
• To investigate the impact of measurement error on divergence estimates.

Main Methods:

• Analysis of gene expression data from mouse, rat, and human species.
• Application of the correlation coefficient to gene expression profiles.
• Comparison of correlation coefficient results with Euclidean distance for uniformly expressed genes.

Main Results:

• The correlation coefficient shows artificially high divergence for genes with conserved uniform expression patterns.
• Genes with conserved uniform expression exhibit significantly higher divergence when measured by correlation coefficient.
• Euclidean distance provides lower, more accurate estimates of expression divergence for these genes.

Conclusions:

• The correlation coefficient may be unreliable for assessing divergence in uniformly expressed genes.
• Euclidean distance is a more suitable metric for evaluating expression divergence in conserved, uniformly expressed genes.
• Careful selection of statistical methods is essential for accurate comparative genomics.