Alternative to Cufflinks ?

Question

[Rox]

Hi everyone,

This post is the second part of a question I already ask and I couldn't find any answers :

C: Cufflinks is frozen

I need a alternative to Cufflinks in order to process the bam files produced by Tophat2. Indeed, I tried several time to use it, but each time, Cufflinks froze at the exact same location on my canu assembly (tig00000382:15043-15157) , and it can last forever.

My RNAseq data seem to be fine, since a colleague could analyse them (with an other assembly reference to align against with tophat), and use Cufflinks who didn't behave the same way, it didn't froze and do the job.

You can have more information about my problem on an issue I posted on github : https://github.com/cole-trapnell-lab/cufflinks/issues/72

It seems that I'm not the only one you got this problem.

For an alternative, I heard about Trinity, but I don't know that much about it, and about the differences between Trinity and Cufflinks.

Can I have some advices please ?

Thanks for helping !

Cheers,

Roxane

[GenoMax]

Use featureCounts/HTSeq-count to get the counts of genes from the bam files produced by TopHat. Then onwards with DESeq2/edgeR.

Trinity is for assembling a de novo transcriptome starting with original reads.

[Chris Fields]

Actually, Trinity can do both (de novo and genome-guided assembly):

https://github.com/trinityrnaseq/trinityrnaseq/wiki/Genome-Guided-Trinity-Transcriptome-Assembly

[VHahaut]

Or instead of DESeq2/edgeR you can give a try to limma/voom.

https://bioconductor.org/packages/release/bioc/html/limma.html

Those three differential expression analysis pipeline are really well explained.

[Devon Ryan]

What's the actual goal of the analysis? If you're not trying to find novel genes/isoforms then skip cufflinks. If you are, then try stringTie instead.

[Rox]

Some clarifications about my goals : I'm aiming to fully assemble and annotate the Drosophila suzukii genome.

I need to use Cufflinks in order to have a transcriptome like file of our D. suzukii RNAseq data in order to give the annotation pipeline maker some hints. So I need to use cufflinks to produce a GFF on my own assembly.

I'm not interested in the quantification part for the moment.

[GenoMax]

There is a v.1.0 release of D. suzuki genome/transcriptome available. Have you compared your results to that version?

Answer 1

[i.sudbery]

StringTie is pretty much a complete replacement for Cufflinks for transcript assembly and quantification. Its way faster than Cufflinks (e.g. a recent job that took Cufflinks 40 CPU hours took StringTie 1 CPU hour).

If you want to create your own transcript annotations, I definitely recommend using it.

For quantitaition, I'd recommend the alignment free algorithms (Kallisto, Salmon etc), as several people's benchmarks have recently shown them to be more accurate than the alignment based methods. See here, here and here.

[Rox]

Thanks for the suggestions ! I never heard about StringTie, but I already used Kallisto for transcript quantification and I have to say it's a very good software, it's very fast.

I think I'm going to have a look on StringTie soon.

So you think that Trinity is not a good idea ?

[i.sudbery]

Not at all, in fact, if the idea is to annotate a new genome, you might like to do both and then merge the results with either stringtie --merge or cuffmerge.

[Rox]

Thanks for the tip ! So just to clarify, you think I might use both String Tie, Trinity and then merge the two files with Cuffmerge so I get better results for my annotation, right ?

[i.sudbery]

Yeah, thats what I was thinking.