We have some questions about cufflinks denovo assembly which we are grappling with. We could not do denovo assembly on our data because cufflinks ran out of memory we used 600gb per sample, the maximum in our cluster). But we could get an assembly from one of our samples which had a much lower file size (about 1/5th sequencing depth). We were wondering if we could split the bam files from the other samples into 5-10 chunks and denovo assemble them individually. We anyway make a cuffmerge at the end. Will we lose sensitivity for genes with low expression levels if we do this?
Also, we we thinking if we could merge a standard annotation (gencodem4 is what we are using) to our denovo assembly? As far as we understand denovo assembly is not perfect and we could lose some true transcripts in the process.