Is rsubread splice-aware? Should I use HISAT2, rsubread, or some other aligner?

Question

[rezaeir75]

• I'm currently learning different approaches for RNA-seq data QC, alignment, and analysis. However, I'm confused that which tool is best for alignment. Because I'm comfortable with R, I'd rather stay in R environment and use Rsubread but I want to be sure that I'm not sacrificing accuracy for ease.
• What do you think about Rsubread? How does it compare to other tools considering speed and accuracy? Is it splice-aware? How does it compare to HISAT2?

[Kristoffer Vitting-Seerup]

What is the end goal of your RNAseq analysis? The answer probably depends on that :-)

[rezaeir75]

my primary objective is to find differentially expressed genes in tumor cells before and after treatment. but I also want to search for miRNA-mRNA regulatory networks as a secondary goal.

Answer 1

[Kristoffer Vitting-Seerup]

I wrote something about quantification considerations in the vignette of my R package IsoformSwitchAnalyzeR which might be relevant. TL;DR: you probably need to venture outside of R but those tools are really easy to use :-)

[rezaeir75]

Thank you for your answer. I finally went with kallisto and pseudo-alignment approach. I'm learning to use your package for DTU. However, there is a problem. Maybe this is not the right place and I should ask another question. But anyway, the problem is that when I run importRdata() with the following arguments:

mySwitchList <- importRdata(
  isoformCountMatrix   = Txi_trans$counts,
  isoformRepExpression = Txi_trans$abundance,
  designMatrix         = targets.mod,
  removeNonConvensionalChr = TRUE,
  addAnnotatedORFs=TRUE,
  isoformExonAnnoation = "Homo_sapiens.GRCh38.100.chr_patch_hapl_scaff.gtf.gz",
  isoformNtFasta       = "Homo_sapiens.GRCh38.cdna.all.fa.gz",
  showProgress = TRUE
)

it gives me this error that:

Step 1 of 6: Checking data...
Step 2 of 6: Obtaining annotation...
    importing GTF (this may take a while)
Error in importRdata(isoformCountMatrix = Txi_trans$counts, isoformRepExpression = Txi_trans$abundance,  : 
  The annotation and quantification (count/abundance matrix and isoform annotation) seems to be different (Jaccard similarity < 0.925). 
Either isforoms found in the annotation are not quantifed or vise versa. 
Specifically:
 172247 isoforms were quantified.
 226798 isoforms are annotated.
 Only 170828 overlap.
 1419 isoforms quantifed isoforms had no corresponding annoation

This combination cannot be analyzed since it will cause discrepencies between quantification and annotation thereby skewing all analysis.

based on the rest of this output and its recommendation to use ignoreAfterPeriod, I adjusted my code and I got the same output except the line:

 1419 isoforms quantifed isoforms had no corresponding annoation

became

 214 isoforms quantifed isoforms had no corresponding annoation

so, my question is, how should I fix this issue? Thank you for your help and this awesome package.

[Kristoffer Vitting-Seerup]

Kallisto sounds like a good approach. With regards to IsoformSwitchAnalyzeR It is not appropriate to ask a new question as a comment (since it makes it very hard for other people to find/navigate the info/solutions). Either ask it as a new question here on biostars or in the IsoformSwitchAnalyzeR google group (where you can also look if other people have had this problem).

[rezaeir75]

Thank you for your help. I found my answer here.
Thank you for your awesome package

Answer 2

[Gordon Smyth]

Rsubread and HISAT2 are both splice aware. All the RNA-seq aligners are splice-aware, otherwise they wouldn't be RNA-seq aligners.

I personally find Rsubread the fastest and easiest aligner for a gene-level differential expression analysis (but note I'm an Rsubread author).