tweeDEseq

tweeDEseq models RNA-seq count data using the Poisson-Tweedie family to perform differential expression analysis that accommodates heavy tails and zero-inflation not captured by Poisson or negative binomial models.

Key Features:

• Poisson-Tweedie family: Utilizes the Poisson-Tweedie family of count distributions to capture heavy-tails and zero-inflation without altering configuration parameters.
• Flexibility and accuracy: Handles large-scale RNA-seq datasets with extensive biological replication by accommodating diverse count distributions to better fit underlying biological variability.
• Improved performance: Simulation studies on synthetic and real RNA-seq datasets show it yields P-values that are equal to or more accurate than those produced by competing methods across configurations.
• Reproducibility: Detects differentially expressed genes with improved performance and reproducibility, evidenced by studies on sex-specific gene expression changes in human lymphoblastoid cell lines.
• Comparison with microarrays: Has been validated against microarray results to confirm reproducibility of differential expression findings.

Scientific Applications:

• Differential expression with many replicates: Differential expression analysis of RNA-seq datasets containing numerous biological replicates.
• Modeling complex count distributions: Modeling RNA-seq count data exhibiting heavy tails and zero-inflation where Poisson or negative binomial assumptions fail.
• Sex-specific expression studies: Investigating subtle sex-specific gene expression changes in human lymphoblastoid cell lines.
• Cross-platform validation: Benchmarking and validating RNA-seq differential expression results against microarray data.

Methodology:

Fits RNA-seq count data to the Poisson-Tweedie distribution family to estimate count distributions and enable comparison between groups of samples.