Figures generated from RNAseq

Question

[kcavagnero]

Hi all,

I want to preface this question by saying that I am a complete noob when it comes to rna seq data analysis and bioinformatics in general. I had some sequencing done by my university's core for 8 samples - 4 conditions with 2 replicates each, and I would like to show with figures the similarities between the four conditions and also the differences. Does anyone have any input on the best way to do this? I am thinking a PCA plot, heat map with all of the genes that are similar, and a heat map with all that are different. Does this sound like a reasonable approach? As far as I am concerned, heat maps would only show the fold change, but I think it would also be informative to show the absolute expression -- I am wondering if maybe not heat plots and instead use simple bar graphs with TPM values? Lastly, I am in a bit of a rush to get some figures made for a manuscript and if you have any recommendations as to how to do this in a quick, cost-effective, user-friendly web-based manner, please let me know. Any help would be much appreciated!

Thank you in advance!

Kellen

[cpad0112]

Follow griffith's tutorial on RNAseq data analysis: https://github.com/griffithlab/rnaseq_tutorial. Not web based, but easy to follow. Featurecounts-DEseq2 tutorial can be followed from the blog: https://digibio.blogspot.com/2017/11/rna-seq-analysis-hisat2-featurecounts.html

[Friederike]

are you comfortable with using R?

[kcavagnero]

Nah. Trying to learn now, but not sure I'll have time as my boss wants this out asap.

[Friederike]

I guess, then you should follow Lior Pachter's advice on how to write a paper in five minutes highlighting the Maya'an's Lab BioJupies Notebooks. If your experiment doesn't have a very complex experimental design, that may help.

Answer 1

[Friederike]

Usually, you would start from a matrix of read counts (integers). If you import those into R, there are numerous packages that will allow you to achieve the kinds of plots you envision (and more).

Generally, you would:

1. normalize the counts for differences in sequencing depth between the samples
2. possibly account for the dependence of the variance on the mean, e.g. using DESeq2's rlog function

The easiest way to generate the plots you're after would then be the pcaExplorer package.

For more detailed code and explanations, you can look at Chapter 5 of our course material.

For determining logFC, you should definitely make use of the DESeq2 (or limma or edgeR) package as they will try to model the gene counts with fairly sophisticated approaches to get fairly robust results (that would be chapter 6 in the course notes).