Low transrate assembly score with good bowtie2 mapping?

Question

[fgrazom]

Hi Everyone!, this is my first question here, but i have been following you and getting tips from this community for a long time.

So, up to my question, i got data from 8 samples of a paired true-seq RNA-Seq experiment, i used sortmeRNA to delete rRNA, then used trimmomatic to clip adaptors, trim reads with a slidingwindow and removed all the reads below 50 nt. Then we used Trinity with default options (just as a first try), and got an assembled transcriptome with around 85k contigs, everything seem to be alright until i tried transrate to evaluate the assembly, it outputs an incredibly low score (0.04) and this seems to be caused because most of the reads are not aligning to the transcriptome, so i ran again trinity (i noticed i didn`t specify ss_libtype) specifying the ss_libtype and got another assembly that got a transrate score like the former.

I´ve been reading and looking for ways to improve the assembly, I was about to start all over again with the QC and then i tried running an alignment with bowtie2 to the first assembly using one of the samples and it turns out i got around 90% of the reads aligned to the assembly.

So my question here is, has anybody been in this type of situations?, and what would you do?, should i trust transrate or bowtie2?. i was thinking about trying to assemble with different kmer size and choose the best assembly, but now im stuck with this.

Edit: I'm using the first assembly, because i was sure it was strand specific but looking at the protocol used i noticed i was wrong, so i discarded the second assembly.

[h.mon]

until i tried transrate to evaluate the assembly, it outputs an incredibly low score (0.04) and this seems to be caused because most of the reads are not aligning to the transcriptome

And what is the mapping rate? Did you map the raw reads, or the trimmed, rRNA cleaned reads?

[fgrazom]

Hi!, the mapping rate is just 0.2, mapping the trimmed rRNA cleaned reads.

[h.mon]

It is a known issue (Mapping of TransRate using SNAP, SNAP failing in v 1.0.3, SNAP failed with Transrate v 1.0.3), without an answer from the developers still unresolved.

You could try install SNAP manually, and check if it runs properly.

[fgrazom]

i didn`t know that, i thought it was using salmon instead of SNAP.

I'll try installing SNAP and let you know the outcome.

Thank you very much!,

[jamie.adkins39]

Hi there,

Did this resolve the issue? I'm running into the same issue with my assemblies currently. When I map my reads back using just salmon, I have very high mapping rates (around 90%) but when running transrate my p mapped is only 0.03 thus my score is very low. I haven't had any luck resolving this issue on my own, and I'm not actually getting any error messages.

[MBarcelo]

Hi! I had the same issue. I installed transrate from bioconda and gave me a very bad mapping rate and overall transrate score (compared to the good mapping from bowtie2). However, for other purposes I was using oyster river protocol, that runs three different assemblers and then compares them with transrate within a single pipeline. The output from the transrate build within oyster river protocol was giving me good results. So I guess there is a problem with the available stand alone version of transrate. It is far from ideal but if you download oyster river protocol you can run the transrate version that is build within this program and it works... Hope it helps!