Calculating isoforms abundance using RNAseq data

Question

[iditk]

good morning,

I'm interested in calculating isoform abundance in a certain tissue, and be able to compare the abundance between human patients / mouse.

From what I can tell my best option is : (1) to use methods (such as RSEM) that quantify the isoforms using known annotations of the genome and isoforms, because both genomes should be well annotated. (2) then use a method based on splicing events (such as MISO) or isoforms (such as EBSeq/cuffdiff) to compare the abundance. In this step both options looks problematic. Any recommendation for such a method, that will work well with quantification methods? or other thoughts? I have access to raw data if it matters.

Thanks!

Idit

[Sam]

Have you trying using MATS? Or if you don't mind having alternative exon expression instead of alternative isoforms, then you can try DEXSeq

[iditk]

haven't tried DEXseq or MATS yet, but I would prefer working with isoforms, not alternative exons.

[Manvendra Singh]

Each exon would be annotated by its gene name

so beyond all the isoforms which are already there in reference gtf, if you get alternative exon usage then that would denote alternate isoform also

Answer 1

[Mikael Huss]

You could try Kallisto for quantification and Sleuth for DE analysis. Like CuffDiff or RSEM/ebSeq, Sleuth can do proper DE testing on isoform abundance, but it also allows using covariates in a linear model, like limma/DESeq/edgeR. It's also possible to use Sleuth with Sailfish or Salmon if you prefer that.

Answer 2

[Manvendra Singh]

If you are comparing human and mouse then first step would be to take only those RNA-seq reads for mapping on their respective genomes which are uniquely mappable on both genomes.

because for the cross-species analysis, we should be owing to knowledge that there are so many different genomic insertions and deletions between every species.

once you have mapped the reads in this way, then I would do splicing detection manually in following way, because I personally am disappointed with MATS and MISO, (someone please correct me if I am wrong):

1. modify the gtf file in such a way that each exon of the gene would be represented with different names.

   e.g geneA_exon1
   geneA_exon2 and so on

   another geneB
   geneB_exon1
   geneB_exon2 and so on.

now, I would run HT-seq on mapped files providing the modified gtf.

once we have got read counts over each exon, we calculate psi (Reads In/Reads out) for each exon, its like measurement of the reads mapped on exon of interest and reads skipped out to different exon., you can keep threshold of number of reads and foldchange of psi during comparison which would depend on how stringent you want to make this pipeline.

finally level of psi would tell you differential exon usage or differential isoform abundances

HTH

[iditk]

Thanks for your answer HTH, I wasn't clear enough. I would work only on human samples or mouse samples (and won't be comparing it).

Do you run the manual procedure on each exon in a gene? To your knowledge, non of the software available can calculate PSI properly?

[Manvendra Singh]

Dear iditk, HTH means "Hope This Helps". My name is Manvendra Singh "Manu" as its written in profile.

I suggested this method because, this way you know what you are doing in each step to calculate psi, where you can play with thresholds to get results less but high confident ones.

No I am not saying that, All tools would be good.

MATS gave me result for 1% of total number of exons in gtf file

One drawback of MISO is that it lacks statistical methods for handling groups of samples.

DEX-seq is better in this case but after validation they got 27 out of 40 detected genes as true positives.

Some tools consider only reads mapping on exon or tarnscript, some consider both the junction reads and exon reads, power of statistics is good in all the tools.

so its on you which one you want to use for your purpose.

Answer 3

[EagleEye]

Did you try GESS with MISO which I have explained in other post? How to determine alternative splicing read counts

Answer 4

[derseb]

Try to calculate the PSI metric for all the exons in your gene and then compare the splicing status of exons across species:

http://onlinelibrary.wiley.com/doi/10.1002/0471142905.hg1116s87/abstract