This is a test version of Biostars. For the public version, visit https://www.biostars.org.
3 patients with 2 samples collected pre and post clinical intervention

The cohort contains 3 patients who each had samples collected before and after an intervention. For each patient, two samples before (1 from tissue A and 1 from tissue B), and two samples from each tissue after the intervention.

Sample meta data: 

meta <- data.frame(Patient = c(1,1,1,1,2,2,2,2,3,3,3,3), Tissue = c('A','B','A','B','A','B','A','B','A','B','A','B'), Treatment = c('Pre','Pre','Post','Post','Pre','Pre','Post','Post','Pre','Pre','Post','Post')

meta

Patient Tissue  Treatment
    1   A   Pre
    1   B   Pre
    1   A   Post
    1   B   Post
    2   A   Pre
    2   B   Pre
    2   A   Post
    2   B   Post
    3   A   Pre
    3   B   Pre
    3   A   Post
    3   B   Post

I have been using the Sleuth package and performed the following model for each tissue: ~Treatment + Patient. No hits - likely due to low sample size.

It was suggested to me by a colleague to look at each patient and tissue individually. For example, for tissue A, calculate the fold change for pre and post. Is this appropriate? What other analysis designs make sense for these data?

Thanks in advance.

rna-seq sleuth study design
• 998 views • link
updated by Kevin Blighe Kevin Blighe • written by lmmjrf • lmmjrf

1 answer

The situation in which you would stratify your analysis by tissue is where the tissue-specific differences are large / 'appreciable' such that these differences result in the biasing of the normalisation procedure, and also the statistical inferences that you make in reference to treatment. By 'stratify' here, I mean to literally normalise tissue A and B separately.

You can check for evidence of tissue-specific differences by processing all data together, initially, and then generating a PCA bi-plot. If you then see the samples separating along PC1, with PC1 contributing toward a large proportion of explained variation, then you will be better informed.

If you still wish to proceed as normal with all samples combined, then it may be easier to create a merged factor for your metadata comprising:

meta$merged_levels <- paste0(meta$Tissue, '_', meta$Treatment)

Then, your design formula would just be:

~ merged_levels

Kevin

• 0 views • link
written by Kevin Blighe Kevin Blighe

Log in to answer this question.

More posts like this