So I have a genome where many gene models are broken or incomplete, and where the TSS of most genes is likely wrong. I’ve been kicking around doing long-read RNAseq to improve these annotations. One thing I’m struggling with is how to get the most out of what promises to be an expensive experiment. Ideally, I would get some reads for every protein coding gene in the genome. Of course, there’s no context where all genes are expressed at once. One option would be to sequence multiple samples and hope to retrieve most genes by choosing samples that, in aggregate, express most genes. The number of samples involved makes this expensive.

Another option might be to sequence a few samples that will be interesting for some other biological question and accept that whole coverage of protein coding genes will be incomplete, it is still going to represent a big improvement.

Any of that would be following polyA enrichment and/or rRNA depletion.

But my ideal experiment would be to sequence a pooled sample that contains roughly equal amounts of all transcripts. I am kicking around an idea for making such a sample, but am not sure if this is, for one reason or another, a dumb idea (or something someone else has done where I could be spared the headache of method development), so I’ve come to ask the hive mind.

Could I enrich a pooled sample of RNA from many tissues using a biotinylated randomer? Would this generate the kind of diversity that I’m thinking would be ideal for my sequencing library? Or is this just doomed to fail?

I was thinking of using something like a biotinylated random octomer for pulldown, then proceeding from that to library prep for sequencing on pacbio revio.

Open to alternate suggestions also. Really just aiming to maximize the bang for my buck where using long read sequencing is concerned.