I am (VERY) new to NGS/RNAseq/Transcriptomics and am trying to determine the best options for bulk sequencing samples from a recent study. We have a 2x2x2 study design (genotype x sex x treatment), using adult mice in a knockout transgenic mouse model. For sequencing, we pulled down a single cell type from a single brain region using Miltenyi Biotec nanobeads. From that, we extracted the RNA using a Qiagen kit. We eluted a total volume of 27ul per sample and sent 3ul of each sample offsite for analysis. Qubit concentrations are 2.0 - 6.1ng/ul. Average is around 3.1ng/ul, however, we had 5 samples below 2ng/ul on the Qubit, and they tell us they don't trust the concentrations that come out of the Agilent FA, which are wildly different from the Qubit numbers).

Based on RQN, we have selected the best 3 biological replicates per group (24 samples total) for potential sequencing (outsourced), Unfortunately, we have to use 3 of the low-concentration samples, and I fear I may have to use too much of my limited extracts (none left for follow-on work) for those samples. (The organization we are looking at outsourcing to doesn't offer preamplification services... But I digress...that's ANOTHER issue...)

At the moment, I am trying to confirm which options are best for our study (cost/benefit ratio), and would really appreciate feedback on whether or not ERCC-spike ins (and which kind) are worth the added expense.

We are looking at bulk sequencing (Nextseq 500) 75bp PE, 50M reads. We are expecting/hypothesizing that we will find upregulation and downregulation of a subset of transcripts based on genotype, treatment and sex. However, to the best of our knowledge, RNAseq and our treatment has not been performed on this transgenic before, so this is an early stage hypothesis. This (and the complexity of our design) is why we are NOT doing the much more expensive scRNAseq.

From my limited understanding, it sounds like if we opt for spike-ins, we'd really want the ExFold ERCC spike-in (which contains both mix1 and mix2)? This is very expensive and adds another $3300 to our costs. I have a friend who does bioinformatics in the cancer research world, and he said,

"Regarding the ERCC spike-ins, as I said, unless you're expecting global changes to transcription (i.e. an overall higher level (or lower level) of transcription, the spike-ins probably aren't going to buy you much.  DE analysis software like DESeq2 does its own normalization, and the manual specifically advises you *not* to normalize first, as you might with the spike-ins.  Having said all that, I don't have personal experience with analyzing RNA-seq with spike-ins; somebody who does might tell you the opposite story. "

I don't think that we are expecting to see global overall changes in transcription (though it is suspicious that our lowest concentration samples are from MALES in one treatment group....).

When I looked for answers here around spike-ins, the most recent questions in RG were from 2016. I'm wondering if things have changed since then in terms of software normalization, and if there are any other considerations we might have given our study paradigm.

Any and all advice/recommendations are welcome!

Thanks,

-Natalie