Hi Michael,

I want to pass a corrected model matrix to DEseq following the collapse of some technical replicates and I just wanted to make sure I am doing things in the correct order and grabbing the right data.

First I created an arbitrary dds object with a design including only 2 of the 3 elements of my design. Then I collapsed the technical replicates. Then filtered the collapsed counts, followed by the creation of the nested model. Finally I re-created the DESeqDataSetFromMatrix object and I hope, took the counts and colData form the correct objects?

   ##############################################
    ###create a dds object from transcript counts 
    #############################################
    #Create a deseq object
    dds <- DESeqDataSetFromMatrix(cts,
                                  colData = finalQA,
                                  design = ~ women + serum)

    #################################
    #Collapsing Technical Replicates
    #################################
    #we want to collapse technical replicates
    collapseTrial <- collapseReplicates(dds, groupby = finalQA$name, finalQA$shortName)

    #########################
    #Filter the Deseq2 object
    #########################
    #filter any row less than 20 counts in at least two samples
    filter <- rowSums(counts(collapseTrial) >= 20) >= 2

    collapsedDDS <- collapseTrial[filter, ]

    #now make a new item for the model matrix 
    collapsedDDS$nested <- factor(rep(rep(1:5, each = 4), 2))

    #############
    #the new model 
    #############   
    model <- model.matrix(~ infection + infection:nested + infection:serum, colData(collapsedDDS))

    #Re-run the DESeqDataSetFromMatrix 
    #access counts and colData 
    collapsedDDS_nested <- DESeqDataSetFromMatrix(counts(collapsedDDS), 
                                                  colData = colData(collapsedDDS), 
                                                  design = model)
   ########### 
   #Run deseq
   ###########
    deseq.dds2 <- DESeq(collapsedDDS_nested)

Thank You!!

Theres no need to create a second dds. The one after collapsing is fine, and you can replace the design with:

design(dds) <- ...