Hey everyone,
I have some samples of RNASeq that had different approaches in sequencing (sequenced in 4 lanes vs sequenced in 12 lanes, separated in 3 runs, 4 lanes per run). I'm gonna call the first ones deep sequencing and the latter ones repeated shallow sequencing.
After aligning and getting the reads per gene, samples from the same group cluster separately according to which one of the two strategies were used - this happens in several groups.
Following that, I would like to ask two questions: 1) In my view, sequencing with high depth is different that sequencing rather shallow but repeated. The patterns associated with low or even some medium expressed genes are better caught with the first strategy, while the latter will tend to catch very highly, highly and some medium expressed genes (depending on the strategy). This will lead to differences among samples from the same group based on the strategy. Does this make sense? 2) Does anyone know somewhere someone has performed any kind of benchmarking regarding these strategies or similar ones?
Thanks in advance
You can pool your repeated shallow sequencing to generate deep sequencing data (esp if they're from the same library). Furthermore, you can use the differences between each of the shallow sequencing runs as technical replicates and use that in your data normalisation strategy.
Can you provide the ordinations you have based your inference on (maybe including some additional PCs)? Can you also provide more information about the samples and RNAseq prep? What is the difference between the deep sequencing and shallow sequencing samples? Are they technical replicates of the same sample?
Your takeaway about depth vs what you can tell seems right to me, but this is pretty straight forward logic to follow. The more you sequence, the more of the lower expressed expressed genes are accurately estimated. How would this affect your analyses? In most cases for DE analysis, it shouldn't, as large fold changes will still be detected regardless of how accurate the actual TPM is since you lose accuracy around lowly expressed genes.