I have a couple of TruSeq/HiScanSQ mRNA-seq from blood samples (no globin depletion) from subjects with different rare autosomal recessive disorders related to nucleotide-excision repair or junction resolution. Of course, we have few biological replicates of each condition (e. g. only two with Bloom syndrome). Despite being associated with a specific gene, nobody investigated their expression. So, we´re not even certain if we could group samples with the same alleged condition. Given this scenario, I want to use DESeq2 to find out differentially expressed genes. The problem is: few biological replicates in the case group and possibly cases within the same condition with very different values of expression. I know that in edgeR one could 'fake' some parameters. Is there a way of doing the same in DESeq2?
How many samples total and how many groups? Were the samples run in batches? Analysis with very few replicates, even 2 per group, is still possible with DESeq2, and I'd recommend this rather than inserting guesses for parameters (which parameters are you referring to?)
For now, as a pilot, we have 3 healthy controls and 2 cases of Bloom's. The focal gene (BLM) is somewhat responsible for chromosomal integrity. Controls and one case were run on the same lane. The other case on other lane/cell.
You can certainly do an exploratory analysis with design
~ conditionor~ batch + condition. The first one leaves you with 3 degrees of freedom for estimating dispersion, and the second one with 2 degrees. Having one of the two cases on a different lane than the controls is not ideal: you would typically want to evenly split cases and controls across the different batches to give some hope to separate the batch effect from the condition effect.I´m aware of such design issues. As usual, I came into the project way after experimental runs. So, I'm trying to do my best and asking for some expert advice. I hope to convince PIs to (re)design things before next round of runs.