How to get the longest fasta sequence including all possible switching isoforms of a gene out of isoforms

Question

1. I am working with RNA-seq data without a reference genome/transcriptome and am instead using a Trinity de novo transcriptome assembly.
2. I analyzed both isoform and gene expression abundances using RSEM.

3. And got differentially expressed GENES via DESeq2. In the next step I want to get fasta sequences of differentially expressed genes to use them in the subsequent BlastX/Blast2Go analyses.

   Considering the fact that my Trinity created assembly is based on isoform id:

   How can I extract the gene sequence out of it? Is that a right thing to combine all the isoforms of a single gene in order to generate the longest possible isoform assembly?

   If so, how can I do it to have a union-exon (combined form of all the exons present in the isoforms of an individual gene) sequence used in my Blast analysis?

   If not, how should I proceed with the data type available? Trinity assembled file includes the isoform column as:

   TRINITY_DN1780_c0_g1_i2 
   TRINITY_DN1780_c0_g1_i1 
   TRINITY_DN1780_c0_g1_i3 
   TRINITY_DN1780_c0_g1_i4 
   TRINITY_DN1780_c0_g1_i5
   

   and the DESeq2 produces the output (gene differentials) as:

   TRINITY_DN1780_c1_g1
   

Answer 1

From your question it isn't clear if you are trying to identify the isoforms that have a very high or very low differential expression value, or if you are trying to correctly find an orthlog for your top hits using BLAST.

If your question is about isoform-level quantification, identifying DE at the isoform level is challenging and DESeq2 is only designed for gene-level analyses. That is why it is only outputting genes rather than a specific isoform.

Answer 2

There's no clear way to go from a trinity "gene" (this is just a cluster of isoforms with shared sequence) to a linear gene sequence that will be blastable and not cause a number of headaches (e.g., due to swapping the order of exons).

What I would suggest that you do is either Blast2GO the isoforms within the Trinity-defined "genes" or take the highest expressed isoform within a "gene" as the sequence you feed into Blast2GO for that gene. Alternatively just use isoforms for everything and avoid this whole issue.

Answer 3

There are a lot of questions in this post, and the main ones have been answered. Here is an answer to the one that has not been answered yet:

How can I do it to have a union-exon (combined form of all the exons present in the isoforms of an individual gene) sequence used in my Blast analysis?

I don't think you should do this, but if you want one of the usually many plausible linearized representations of transcript isoforms, then Lace would do the trick.

Answer 4

It appears from the comment in another question that this question is also about transcript-level differential expression analysis.

If instead of using RSEM you use kallisto for transcript quantification, then you could use sleuth to perform transcript-level differential expression analysis.