Is differential gene expression coupled with differential transcript usage more informative than just differential transcript expression?

Question

[peter.brockport]

Hello!

I'm concerned that I may be taking the wrong approach to this problem. I'm trying to identify transcripts important for day 0 and day 2 in the differentiation process of a particular cell type. I've assembled and quantified transcripts from RNA-Seq data using STAR and Stringtie. So, now I have data that can be described pictorially as in the attached picture. (This is a simplified description) In the picture, the green rectangles are genes, and the squiggles below them are their respective transcripts and count information.

Data representation - > https://imgur.com/gallery/bPFioZm

Now, keep in mind I'm a beginner. So, I may be wrong on some of the intricacies! Anyway, here is my understanding of the following analysis strategies.

Differential Gene Expression (DGE) analysis tells us that genes 1 and 3 were differentially expressed between day 0 and day 2​. A popular tool for this kind of analysis is DESeq.

Differential Transcript Usage (DTU) analysis tells us that on day 2 in gene 3, transcript E was differentially expressed. A popular tool for this kind of analysis is DRIMSeq.

Differential Transcript Expression (DTE) analysis tells us that transcripts A,B,C, and E were differentially expressed between day 0 and day 2.

I am currently using DESeq and DRIMSeq to examine DGE and DTU in my data by following the workflow from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6178912/ However, I'm wondering if I really just need to examine DTE in my data to answer the question "What transcripts are important for the differentiation process between day 0 and day 2?" so my question is: does DGE and DTU combined provide the same information as DTE? If not, what tool can I use to analyze DTE in my data?

Thanks in advance for the help.

EDIT: to be more specific about my goal I have two aims:

• I'm looking for changes in lncRNA expression from day 0 to day 2 of adipogenesis
• and also looking for novel lncRNA transcripts.

I'm predicting that there are transcript level changes (lncRNAs) that are driving my process. For that reason I use de novo assembly with guidance from an annotation. Next, filter out any transcripts that arent lncRNAs by comparing to Gencode lncRNA GTF. Then I quantify the filtered data with stringtie -e The point I'm currently at is analyzing the quantification data for differential expression.

Answer 1

[rpolicastro]

In regards to the workflow steps, are you sure there is no transcriptome for your organism already?

Also, the question about differential transcripts versus genes is somewhat dependent on which level you think can answer your question most appropriately. Do you think it's mostly gene level or transcript level changes that are driving your process? For differentiation I would tend to assume large changes at the gene level, so would probably focus my efforts there first, especially if you are new to the field. Standard DESeq2 analysis and quality control is a robust and well documented, so it will give you a good feel of the quality and structure of changes to be expected.

[peter.brockport]

Hey, thank you very much for the response! I'm working with mouse RNA-Seq data, so I do have access to a transcriptome. I went and updated my question with more details to make my goal more clear.

The type of DE I should use is confusing me. I've learned I can use DESeq2 which tells me which genes the lncRNAs originate from have been DE. But I am also curious about the specific isoforms. DRIMSeq tells me information on DTU, and I believe this gives me useful information. But is there a better way to determine which transcripts are DE between days