Assessment of data transformations for model-based clustering of RNA-Seq data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• RNA-Seq data analysis differs significantly from microarray analysis.
• RNA-Seq data often exhibit non-normal distributions (Poisson or negative binomial), unlike microarray data which are typically normal.
• Limited research exists on the impact of data transformations on Gaussian model-based clustering for RNA-Seq data.

Purpose of the Study:

• To investigate the impact of different data transformations on Gaussian model-based clustering performance for RNA-Seq data.
• To assess how these transformations affect the accuracy of estimating the correct number of clusters.
• To evaluate clustering performance using metrics like adjusted rand index, clustering error rate, and concordance index.

Main Methods:

• Simulated RNA-Seq data were generated with varying cluster sizes and numbers.
• Four data transformations were applied: naïve, logarithmic, Blom, and variance stabilizing transformation.
• Gaussian model-based clustering was performed on the transformed data.

Main Results:

• Data transformations generally improved clustering performance compared to no transformation.
• Transformations that made the data distribution more Gaussian-like led to better clustering outcomes.
• The choice of transformation impacted the accuracy in estimating the number of clusters.

Conclusions:

• Data transformations are beneficial for applying Gaussian model-based clustering to RNA-Seq data.
• Selecting appropriate transformations can enhance clustering accuracy and the reliability of cluster number estimation.
• Further research is warranted to optimize transformation strategies for diverse RNA-Seq datasets.