A phylogenetic mixture model for the evolution of gene expression.

Area of Science:

• Genomics
• Evolutionary Biology
• Bioinformatics

Background:

• Microarray platforms enable genome-wide gene expression analysis for comparative inferences.
• Existing methods for detecting natural selection using gene expression variance often fail to account for phylogenetic nonindependence.

Purpose of the Study:

• To introduce a novel mixture model for comparative gene expression analysis that explicitly accounts for phylogenetic associations.
• To disentangle phylogenetic and non-phylogenetic signals, as well as heritable and non-heritable variation.

Main Methods:

• Development of a mixture model for comparative analysis of gene expression variation across multiple taxa.
• Isolation of phylogenetic signal from non-phylogenetic signal and heritable from non-heritable variation.
• Development of a likelihood ratio test for natural selection.

Main Results:

• The proposed mixture model effectively separates phylogenetic and non-phylogenetic components of gene expression variation.
• The model provides a unified framework for estimating across-taxa variance and resolves differences with existing models.
• Simulations and a gene duplication family dataset analysis demonstrate the feasibility, utility, and power of the model and test.

Conclusions:

• The mixture model offers a more accurate and flexible approach to analyzing comparative gene expression data.
• This method enhances the ability to detect natural selection by properly accounting for evolutionary history.
• The developed likelihood ratio test provides a robust statistical tool for inferring selection pressures from gene expression data.