I am doing an RNA-seq experiment to determine differential expression between genes in 4 groups.
My samples are being generated by a protocol that I am testing and I would not like to invest in triplicates or more before I know whether the experiment and RNA-seq will work.
Can I perform an RNAseq with one or two samples from each group and, if the experiment works, run another RNA-seq with the rest of the samples and merge the runs together as triplicates or quadruplicates?
Hi Thiago Leite Knittel
Don't do that, a single sample in each group would tell you nothing if your purpose is to do differential expression. You need multiple samples in each group for the differential expression to be reliable. Merging data from different batches which are generated on different occasions is tricky and likely would confound your results.
I am not sure what you mean by "RNA-Seq will work"!
Alternatively, look for publicly available data that resembles your experiment and use it to guide your design.
Hello Thiago Leite Knittel
It would be better to perform the sequencing of all the samples along with their replicates in a single run to get a better resolution of differential gene expression analysis. Here, I do agree with Mahmoud Ahmed, as by running replicates in different run may leads to higher dispersion among the replicates.
Although it would be better to go with qRT-expression analysis of some publicly available genes of your species or its closely related species before RNA-Seq analysis. If this works then probably your RNA-Seq will also work. However for RNA-Seq analysis, it would be better to run all samples at a time.
Hi,
Technically, you can do that. Just make sure that you sequence at least one sample from each group for separate runs. By doing so, you could check if there is any batch effect between different runs and correct the effect when you want to merge your data in downstream analysis.
However, since your sample size is rather small, it is probably not recommended to do so as mentioned by Mahmoud Ahmed and Romit Seth because it introduces an additional variable into your experiment.
Alternatively, you can go for a pilot sequencing run with fewer samples/groups and less depth. If you think the pilot run gives good results, then you may proceed to an actual run with all samples from every group.
Side note, if your intention is to check your pipeline, you can also download public raw sequencing data and do the analysis with your pipeline.
I hope this helps. Thanks.
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology