I am working on bacterial RNA seq data to perform differential expression analyses.
Considering one bacteria and one antibiotic, the goal is to identify trancripts/genes that are differentially expressed between Resistant (R) and Sensitive (S) strains.
Half of the R/S samples are treated and the other half are untreated.
Treated samples were exposed to antibiotic during different times : 10 min, 30 min and 60 min.
Just few example to clarify the notation :
R+10 means that the sample is a Resistant strain ( R) treated with antibiotic (+) during 10 minutes (10)
S-30 means that the sample is a Sensitive strain (S) untreated (-) during 30 minutes (30)
For each condition (R+10, S-30…) I have triplicates.
I ran 2 different pipelines :
For both workflows, I perform LRT test to compare the full and reduced models.
In the following comparison I want to highlight the effect of the interaction between bacterial status (R or S) and treatment (treated or untreated). To simplify the model I tested separately each timepoint.
Full Model: y ~ RorS + IsTreat + RoS * IsTreat
Reduced Model: y ~ RorS + IsTreat
I am also interesting in the following pattern :
Full Model: y ~ IsTreat
Reduced Model: y ~ Intercept
In this case I consider Sensitive strains only (treatment effect in Sensitive strain)
Results suggest that kallisto/sleuth workflow gives better outputs compared to bowtie2/ featuresCounts/DEseq2 and I want to go further with sleuth.
My concern is the following: I want to improve my analyses pipeline by normalizing all my samples together (R/S, treated/untreated, all timepoint exposure) because I want to be able to compare individual gene expression levels across all samples from all conditions (graphical views...).
My problem is that, for the DE analysis, I don’t know how to subset a sleuth object to consider for example only the sensitive strain samples to test the second model above.
Is it possible to subset a sleuth object to filter on sample of interest? If yes, how? when using sleuth_fit? which part of sleuth object is used in sleuth_fit?
I tried to use functions from dplyr but it does not work on sleuth object.
I already looked for answers on bioinformatic forums but unfortunatly can't find any for now and I am blocked...
Thank you in advance for your help!