Complex designs

DESeq2 also allows for the analysis of complex designs. We heavily encourage trainees to read the entire DESeq2 vignette to learn more:

https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html

Briefly, a few examples may include the need to control for variables, such as a batch effect or biological variable (e.g.: sex, age etc.). The standard process through which this can be performed in DESeq2 is by including the additional variable(s) as part of the design. For example, the following design would test for differences in a variable called condition, while controlling for batch.

design= ~ batch + condition

Note: Does the order of variables matter?

The best practice for DESeq2 is to add your variable of interest (to be tested for DEG analysis), at the last position in the design (as shown in the example above). This order aids in ensuring that default results are pulled from the variable of interest.

Time-series experiments

DESeq2 can be used to analyze time course experiments. This workflow is a great introduction for this type of analysis.

Other complex designs

While DESeq2 is a great tool for many RNA-seq analysis, it also has its limitations. One particular limitation to be aware of is that it does not support differential expression analysis with random effects. A great resource for this type of analysis, is limma (specifically limma-voom). While the limma vignette can be quite large due to its historical role in microarray data, it’s well organized and searchable to direct users to RNA-seq data and it’s limma-voom approach.

See more at: http://bioconductor.org/packages/release/bioc/vignettes/limma/inst/doc/usersguide.pdf

Environment reproducibility

As discussed, secondary analysis with nf-core pipelines already implement what is known as software containers (Docker, Singularity), however tertiary analysis and visualization requires users to build their own custom containers. This can be a critical step in aiding your analysis to be portable and reproducible.

Unfortunately, due to time limitations and workshop scope, we are unable to demonstrate how to build your own containers. To learn more about Docker and Singularity containers, we encourage all trainees to visit our training website for a detailed 3-part tutorial in the “Intro to Docker with RStudio” tab from https://u-bds.github.io/training_guides/.

U-BDS Data Science Office hours

If you have follow-up questions about this workshop or any other data science topic which cover domain that U-BDS cover, please attend our office hours.

It is every Thursday from 1:30pm-2:00. The registration link can be found here.

U-BDS services

If you would like to learn more about the core services please visit our main UAB webpage at: https://www.uab.edu/cores/ircp/bds. If you would like to have a consultation, please fill out the “Enquiry Form” present on the website.