Modeling skewness in human transcriptomes.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Modeling

Background:

• Gene expression data exhibit diverse dispersion patterns due to biological and technical factors.
• Traditional microarray analyses often overlook skewed distribution patterns.
• Standard methods assume symmetric Gaussian distributions, which may not accurately represent transcriptome data.

Purpose of the Study:

• To investigate the distribution patterns of human transcriptomes in publicly available microarray data.
• To develop and validate a flexible mixed model for analyzing non-competitive microarray data with asymmetric and heavy-tailed distributions.
• To address the limitations of standard symmetric models in capturing gene expression variability.

Main Methods:

• Analysis of free-access human transcriptome datasets.
• Development of a mixed-effects model incorporating asymmetric Student's t-distributions for random effects.
• Modeling asymmetry parameters (λ) to capture varying degrees of over-expression or under-expression.

Main Results:

• Significant departures from symmetric Gaussian distributions were observed in probe and differential expression effects.
• The proposed asymmetric Student's t-distribution model demonstrated superior performance compared to standard symmetric approaches.
• Human gene expression data showed notable right-skewed asymmetry in probe effects and both symmetric and left-skewed patterns in differential expression.

Conclusions:

• Skew dispersion patterns are prevalent in human transcriptome data from microarray experiments.
• The developed flexible mixed model effectively addresses asymmetric and heavy-tailed dispersion processes.
• This new analytical approach provides a more accurate method for analyzing human gene expression data, accounting for observed skewness.