I have a bit of problem figuring out the proper way to set dataset and model for DEXeq analysis. We have run gene-wise analysis of RNA-seq with DESeq2 but we also have exon-data that we're trying to analyse with DEXeq.
1) Is it enough for the analysis if I have:
i) Sample data with covariates and additional exon column with "this" and "other"
ii) Countdata with columns for samples that have counts per exon and the sum columns for sums for other exons of the gene per sample. Rows are named ENSG3923903:E001 etc. I created the count matrix manually by calculating the sum columns and then downloading the annotation from ensemble so I could combine it with my data and construct the rownames manually.
2) Couple of questions also regarding the model. In deseq2 we created model.matrices for full and reduced model.
i) is it possible to use model.matrices with DEXeq?
ii) We used formulas :
full = CASE_CONTROL + PRE_MID_POST + CASE_CONTROL:subject.nested + CASE_CONTROL:PRE_MID_POST
reduced = CASE_CONTROL + PRE_MID_POST+ CASE_CONTROL:subject.nested
How can we apply the "exon" variable for the models and run the analysis on DEXeq? We're trying to test whether there DE can be detected between two condition across three time points so that we're accounting for the subject variability.
Help would be really appreciated. (Especially from Alejandro Reyes, Simon Anders, Wolfgang Huber.)