DESeq

DESeq analyzes count data from high-throughput sequencing assays such as RNA-Seq and ChIP-Seq to estimate variance-mean dependence and test for differential expression using a negative binomial model.

Key Features:

• Count data analysis: Operates on quantitative readouts in the form of counts from high-throughput sequencing assays including RNA-Seq and ChIP-Seq.
• Variance-mean estimation: Estimates variance–mean dependence across count data to characterize data variability across the dynamic range.
• Negative binomial testing: Tests for differential expression using a model based on the negative binomial distribution.
• Local regression: Employs local regression techniques to link variance and mean in the count data.
• Overdispersion modeling: Models overdispersion commonly observed in sequencing count data that is not adequately addressed by Poisson models.
• Statistical power improvement: Enhances statistical power by providing accurate variance estimates that reduce false positives due to random variation.
• Implementation: Implemented as an R/Bioconductor package.

Scientific Applications:

• Differential expression analysis: Performs differential expression testing on gene- or feature-level count matrices derived from RNA-Seq and related assays.
• Differential signal detection: Identifies differential signals within high-throughput sequencing datasets, including ChIP-Seq.
• Genomic research: Supports a range of genomic research applications that require modeling and testing of count-based sequencing data.

Methodology:

Estimates variance–mean dependence using local regression and fits a negative binomial model to test for differential expression and to model overdispersion.