Hi,

I am quite new to RNAseq and this is my first analysis and I am not sure whether I did my analysis correctly in my pipeline. I have been searching online to see whether people have similar experiment layout like mine but couldn't really find one, and I have read through the Beginner's guide for DESeq2 and the vignette and also the great rmarkdown file to explain the factor for DESeq2 (https://rstudio-pubs-static.s3.amazonaws.com/329027_593046fb6d7a427da6b2c538caf601e1.html#example-1-two-group-comparison)

Briefly explaining my experimental layout. I am working on a cotton moth, which the experiment involves a lab susceptible line (Gr), and a CRISPR/Cas9 knockout line that was created from Gr, called Cad, the CRISPR/Cas9 line resulted resistance to a toxin. The experiment also involves treatment with a toxin, but only on Cad line, not on Gr.

Therefore, I have 3 groups that were sequenced, 1: Gr (untreated), 2: Cad (untreated) and 3: Cad (treated). I am interested to find out 2 things from RNAseq, GOAL 1: run DGE analysis to see whether the CRISPR/Cas9 knockout would result in any DGE differences, for group Gr (untreated) and Cad (untreated). AND GOAL 2: run DGE analysis to see whether the treatment with the toxin would result in any DGE differences, for group Cad(untreated) and Cad (treated).

My sample info is created as

              phenotype   treatment
GRuntreat1     sus        untreated
GRuntreat2     sus        untreated
GRuntreat3     sus        untreated
CADuntreat1    res        untreated
CADuntreat2    res        untreated
CADuntreat3    res        untreated
CADtreat1      res          treated
CADtreat2      res          treated
CADtreat3      res          treated

My DESeq2 design at the moment is

DESeqDataSetFromMatrix(countData = counts_reverse,
                                      colData = metaData,
                                      design = ~ phenotype + treatment)

However I feel like I have made a mistake here after reading the document https://rstudio-pubs-static.s3.amazonaws.com/329027_593046fb6d7a427da6b2c538caf601e1.html#example-1-two-group-comparison and reading through Michael and other bioinformatician's reply to other people's design questions.

As I am not interested in discovering the interactions between the treatment and the phenotype as Gr was not treated at all. At the moment my design is based on the treatment as the main factor with 2 levels, treated and untreated, and taking the variance in phenotype into account.

However after reading the information online, I am in doubt about my design... I am seeking an suggestion whether I should treat my sequenced groups as two completely independent analysis.

i.e. change my sample info (metadata) and redo the analysis as

DESeqDataSetFromMatrix(countData = counts_reverse,
                                      colData = metaData,
                                      design = ~ phenotype )

For only the Gr untreated and Cad untreated and set the Gr as reference and

DESeqDataSetFromMatrix(countData = counts_reverse,
                                      colData = metaData,
                                      design = ~ treatment )

For only the CAD untreated and CAD treared and set CAD untreated as the reference Thanks for your valuable time and advice!

Cheers, Grace

Different designs are better suited for different questions. What exact question are you trying to answer?