Skip to content

Steps in DESeq function

Aarthi Ramakrishnan
— 1 min read

For any RNA-seq data analysis project, one would most probably end up using the R package DESeq2. The function 'DESeq' is the main function that is called to obtain the differentially expressed genes. But what exactly does this function do?

In short, DESeq function combines the following 3 steps -

  1. estimateSizeFactors - This step calculates the size factors for each sample (explained here).
  2. estimateDispersions - This step obtains the dispersions for each gene. Here, dispersion  does not mean the variance. Rather, it represents the deviation of the variance from the mean. This value is required by DESeq as it assumes a negative binomial distribution, and dispersion is one of the model's parameters. This link has a good explanation on estimating dispersions.
  3. nbinomWaldTest - This step fits the normalized counts data with a negative binomial model and runs a Wald test to find the differentially expressed genes in the dataset.

Bioinformatics

Aarthi Ramakrishnan

Related Posts

Interesting Bioinformatics Articles

Following is a collection of articles which I feel every Bioinformatician must be aware of. I will keep updating this list from time to time - 1. All biology is computational biology 2. Core services: Reward bioinformaticians 3. Importance of stupidity in scientific research

Interesting Bioinformatics Articles

Screen command in UNIX

Screen is a very useful command to have in your toolbox if you frequently use interactive sessions on your supercomputer logged in through a VPN. A VPN typically has a time limit, and you may get disconnected from it without any warning when you have poor internet connection. Screen program

Screen command in UNIX

Types of models in DESeq2

There are 2 major types of regression models one can specify in DESeq2 to explore the raw count matrices from an RNA-seq experiment - * Mean-reference model for Factors * Regression model for Covariates Mean-reference model for Factors - Factors typically represent categorical variable such as Gender, Ethnicity, Race etc. The mean-reference

Types of models in DESeq2