RNA-Seq single time-course experiments -- help building a hypothesis based on cluster plot profiles

Question

[Efejir]

I have RNA-seq data of a replicated single time-course experiment of a particular microbe with 3 time-points. After running the data in R using maSigPro, I obtained some very interesting expression-vs-time cluster plots. You may take a look at the median plots here. You'll see, for instance, cluster 4 showed what would look like a remarkable spike in expression after the 2nd time point. I understand that genes with similar expression profiles are clustered together which would suggest coordinated expression. I'm guessing my next step will be to pick biologically interesting plots and then "annotate these clusters" someway.

My objective is to find "novel" genes I can further investigate (experimentally) as potential drug targets, hopefully based on these cluster diagrams, so my question is how do I analyze these plots to achieve this? Also, are there other analysis I should perform to further support this hypothesis?

I just started RNA-Seq downstream analysis. I've done some reading (workflows and previous papers) and I've found little clear-cut guiding materials on the subject unique to my problem. Slightly overwhelmed, I am hoping for some direction and correction if I'm missing something.

Thanks for the time family.

Efejir

Answer 1

[andrew.j.skelton73]

Rule of thumb that I go by:

Gene level RNA Seq analysis: Alignment (HISAT2, STAR, Tophat2) -> Count (htSeq_Count, RSubRead) -> Differential Expression Modelling with DESeq2.

Transcript Level RNA Seq Analysis: Transcriptome quantification (Salmon or Kallisto) -> Differential Transcript Modelling with Sleuth.

If you have an N = 1 experiment however, you can't perform statistics on your data.

If you're looking for "Novel Stuff", you can use Cufflinks, part of the tuxedo suite, but be warned that often "novel transcripts" don't look real, and you'll need to examine the reads in IGV to see if it looks plausible to you. The only additional validation you could do is look for novel things in your other time point samples too, and look for overlap between the three.

[Efejir]

I'm sorry Andrew, I meant (and have corrected thus) the experiment I'm looking at is replicated. Following up on your response too - thank you. But, don't I get nothing out of the cluster plots?

Answer 2

[ssgmj1991]

Hello,Efejir, i think maybe you can extract the gene list from each cluster, and do GO analysis for the function of them. I think masigpro has pick those signficant change gene from your experiment samples and cluster then according to their time curves. I think from each of the cluster you have the gene list and you may chose the cluster you most interested in to do GO. I hope this may help you:) MG