What shall I do with this RNA-seq data that have no replicates?

Question

[l.burke.1]

I have RNA-seq data from the following two experiments:

1. Onion grown under four different conditions (control, 1, 2, 3), of which three biological replicates were taken from each condition 72 hours after inoculation.
2. Onion grown under condition 2 over the course of 96 hours, with a sample taken at 10 time-points (including 72 hours). However, only two replicates were taken at 24 hours - no other replicates were taken throughout the experiment.

I understand that experiment 1 is fine, but that I won't be able to draw any reliable results from experiment 2 due to the lack of replicates. My plan is to base my research on the first experiment, but is there anything useful I can do with the data from the second? Could I use the second experiment to explore how expression levels of interesting genes identified in ex.1 change over time? It wouldn't form a significant part of the research, but it could help inform future work/form a hypothesis? What be the best way to visualise this? Use GFOLD? Could dispersion estimates from ex.1 be used in ex.2 to act as a proxy for the true dispersion?

Cheers!

Answer 1

[Hussain Ather]

No replicates means your results have limited reliability. You should try to use GFold or DEGSeq. You could also try the NOISeq R package which is designed for no replicates.

[biofalconch]

To add to this, edgeR has multiple ways to deal with no replicates (section 2.11). From the manual:

We do not recommend any of these choices as a satisfactory alternative for biological replication. Rather, they are the best that can be done at the analysis stage...

Answer 2

[Ian]

There might still be useful results in experiment 2, but Pvalue and fold change are only indicative at best. The results might give you leads for further lab experiments, or be impressive enough to warrant a second attempt with replication.