I have multifactor data 1. Participants are healthy population, having both males and females. 2. different excersice intensity for different legs. few participanrts have high intensity on left and low intensity in right, and few have high intensity of right and low in left. 3. samples were taken from both righ and left leg. 4. samples were taken at 4 different time point 5. participants response to training, few participant show high muscle grwoth in high intensity some show high muscle growth in low intensity. I am formulating this to normalize count table i got from htseq and rsem and trying to identify differentially expressed gene responsible for muscle growth and look if there is any biases based on leg, gender, excersice intensity. ddsHTSeq <- DESeqDataSetFromHTSeqCount(sampleTable = sampleTable, directory = ".", design= ~ condition) dds <- DESeq(ddsHTSeq) table_counts_normalized <- counts(dds, normalized=TRUE) samples$Group <- factor(paste0(samples$participants,samples$response, samples$intensity, samples$time, samples$gender)) dds <- DESeqDataSetFromMatrix(countData = table_counts_normalized, colData=samples, design=formula(~Group)) dds <- DESeq(dds)
You make dds from raw counts, not normalized. I'm not even sure what happens if you try to make a dds object with non-integer counts as input.
That method of making a concatenated "group" column is useful for comparing subsets of samples to each other, while using the dispersion of the whole dataset, but I don't think that's what you want to do. You probably do not want to compare one participant at one time with another participant at another time, but that's the kind of comparison you've set up there.
You've also omitted the other key step in setting up DESeq; the call to results, which determines exactly what comparison you are making.