Tophat2 or STAR for IsoformSwitchAnalyzeR analysis

Question

[sbb]

Hello Biostars,

I am doing Isoform switch prediction studies from differential data using IsoformSwitchAnalyzeRhttps://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/isoformswitchanalyzer/SA9sJCQXM54/eNxVps5pDAAJ. The error retrieved :- " you have NOT been doing transcript reconstruction/assembly with Cufflinks/Cuffdiff. If you have not reconstructed transcripts we receomend to use Kallisto or Salmon to do the quantification instead - they are more accurate and have better biase correction methods."

The reason behind seems to be not using Tophat2 --> cufflinks -->cuff merge -->cuffdiff -->IsoformSwitchAnalyzeR, because the pipeline I have followed is STAR -->cuffdiff -->IsoformSwitchAnalyzeR

Should I follow former workflow for Isoform switch analysis or later can also work with some additional efforts? Please comment/suggest.

[Martombo]

better yet would be to use Salmon / Kallisto as suggested by the authors of the R package itself

[Titus]

Maybe i m wrong , but i understood that Salmon / Kallisto are not looking for isoform prediction , they just align against rna ref.

[Martombo]

Yes you are right. If you need to perform isoform discovery / transcriptome assembly the suggested pipeline nowadays is HISAT, StringTie, Ballgown. If you are studying a model organism (eg human, mouse), I would however suggest you to rely on the reference (Ensembl tends to be really extensive) and enjoy the speed and simplicity of Kallisto or Salmon. In any case, using cufflinks is not recommended anymore, as it has been greatly surpassed by the other more recent methods I mentioned.

[Kristoffer Vitting-Seerup]

I would not use Ballgown - I have never seen it perform good in any comparison (and it is just a wrapper using FPKM + Limma). Instead use tximport() and some of the other standard methods (edgR, Limma, DEXSeq2). A nice benchmark of methods (and introduction) can be found in Love et al

[h.mon]

In the post you linked, you had an extremely responsive software author willing to help with your question. Why didn't you open your new question at the same forum? Do you know if the IsoformSwitchAnalyzeR author is an user of BioStars? And why didn't you use IsoformSwitchAnalyzeR as tag, as it would help its author (in case he is subscribed to BioStars) to find posts of his interest?

[sbb]

Well, author @kristoffer.vittingseerup already knows about this post as I created this new post on his suggestion only.

Answer 1

[Kristoffer Vitting-Seerup]

As it is much easier for me to support IsoformSwitchAnalyzeR if all questions are posed in the associated google group I will answer the question there (if posted).