I'm quite new to R and RNA-seq data analysis. I have been looking through some posts here but did not find an answer to what I was looking for, hope someone could help me here:)
I'm using DESeq2 to analyse RNA-seq data set with 96 samples: 16 different species, 2 types of tissue each (M, L), and 3 biological replicates of each tissue of each species. The gene expression level in one tissue type can be completely different from the other, and they don't have to correlate with each other. I'm looking to find differentially expressed genes between species, in either of the tissue types.
I have the counts file from Kallisto and imported the data with tximport, then did an lrt test with all the samples together to generate PCA plot. An outline of the code is as follows:
txi.tsv <- tximport(files, type = "kallisto", tx2gene = tx2gene) dds.int <- DESeqDataSetFromTximport(txi.tsv, SampleTable, design = ~ tissue + species + tissue:species) dds.int <- DESeq(dds.int, test="LRT", reduced = ~tissue + species) vsd.int <- varianceStabilizingTransformation(dds.int, blind=FALSE) plotPCA(vsd.int, intgroup=c("species", "tissue"), returnData=TRUE)
One questions is: is this the right way to do it? As I don't expect any correlation of gene expression in one tissue type to the other, but want to see the difference among species in either of the tissue type. In the PCA plot I do see that the tissue cluster together before species, as expected. And I can generate PCA plot for only one type of tissue like:
vsd.m <- vsd.int[ , vsd.int$tissue == "M"] plotPCA(vsd.m)
If the analyse above is right, now the main question is: how can extract lrt result for all of the species of one type of tissue only? Before looking into the pairwise comparisons, I want to have a list of all the gene that is differentially expressed in any of the species, with an lrt test p-value. Can I simply extract result from the
.int above, or do I need to separate the dataset and run each of the tissue type separately?
I hope I described my questions clearly enough and thank you very much for your help!