I am trying to find differential use of exons across entire genome for two conditions, total number of samples are about 88 (which is a lot for DEXSeq, I am aware of that), I've been trying newest version of DEXSeq_1.26.0 with 50 cores and 1000GB of shared memory, however with days of running I still couldn't get through the estimateDispersions step.
Now I am thinking if I can just do 100 genes per batch to run the program with such setting? and gradually go through all exons batch by batch to identify DEUs? I know DEXSeq share information across exons but I think limiting the shared information among 100 genes maybe a reasonable compromise between accuracy and efficiency? Has anyone tried this way before or have a better solution for large sample and exon number runs with DEXSeq?