Nonfunctioning conversion from .gff to .gtf?

[silviamorins]

Hi,

I am running Nextflow v 19.01.0 with RNASeq data from S. coelicolor; using the Streptomyces_coelicolor_a3_2_.ASM20383v1.37.gff3 file downloaded from Ensembl as annotation leads to the error:

ERROR ~ No such variable: gtf_{makeSTARindex,makeHisatSplicesites}

depending on the aligner I am using.
However, if I execute on my machine the command reported in the respective process here, and then provide the resulting .gtf file at the --gtf flag, the pipeline runs through this step.

I have looked into this process, but have no idea why this is happening.

[lpantano]

Hi @silviamorins

it is indeed weird, thanks for posting. Can you share the .nextflow.log file? maybe we see more information there that can help us.

What is the command line you used to run nextflow?

cheers

[silviamorins]

Hi @lpantano ,

I'm sorry I can't share the whole file because it contains some private information, but here you can find all the lines after the initial description that comes after nexflow.

For running Nextflow I used:

nextflow run nf-core/rnaseq -r 1.3 -profile cfc --fasta 'Streptomyces_coelicolor_a3_2_.ASM20383v1.dna_rm.toplevel.fa' --gff 'Streptomyces_coelicolor_a3_2_.ASM20383v1.43.chromosome.Chromosome.gff' --reads '[...]*_{R1,R2}.fastq.gz' -resume --skip_preseq --reverse_stranded --skip_genebody_coverage

[lpantano]

Thanks, it seems like is not entering that conditional process and the variables then are not defined.

When you get the summary when running the pipeline that is printing all the variables, you see something like:

GFF3 Annotation : Streptomyces_coelicolor_a3_2_.ASM20383v1.43.chromosome.Chromosome.gff

from these lines :

rnaseq/main.nf

Line 254 in 37f260d

if(params.gff) summary['GFF3 Annotation'] = params.gff

[silviamorins]

I see...

I have something else to add, since I've encountered another error further on in the pipeline: featureCounts expects to find an annotation called gene_biotype, which you don't obtain if you run the gffread without the option -F (at least this is what I experienced with the S. coelicolor .gff3 file). Even with the option -F, even if all the information is preserved, the field is called "biotype", so I had to replace it with "gene_biotype" myself in order for the pipeline to run until the end successfully.

[ewels]

so I had to replace it with "gene_biotype" myself in order for the pipeline to run until the end successfully.

Please see the usage docs on how to specify a different field instead of biotype: https://github.com/nf-core/rnaseq/blob/master/docs/usage.md#default-attribute-type

[drpatelh]

@silviamorins Did you manage to resolve this issue?

[silviamorins]

I have converted the .gff3 file to .gtf prior starting the workflow, so I can't say whether performing this conversion inside the pipeline works now (I don't know if any changes have been applied?).

I just realized I had missed @ewels 's comments, sorry - I think his answer would have solved the problem I had with naming the features, but since I had already completed the analysis I haven't used it so far.

[drpatelh]

Ok. Thanks. Is it something you could test quite easily? e.g just running the command that gave you the error with the appropriate modification that @ewels suggested and -r dev? It should give you an error straight away if it's still an issue. If so, I'll try and fix it on the next PR to dev 👍

[silviamorins]

It should be, yes, will do it and let you know!

[silviamorins]

The flag --fcGroupFeaturesType suggested by @ewels worked perfectly for my needs.

The conversion from .gff3 to .gtf still gives me problems, though. With the command:

nextflow run nf-core/rnaseq -r dev -profile cfc --fasta 'Streptomyces_coelicolor_a3_2_.ASM20383v1.dna_rm.toplevel.fa' --gff 'Sequence/Streptomyces_coelicolor_a3_2_.ASM20383v1.43.chromosome.Chromosome.gff3' --reads *_{R1,R2}.fastq.gz' -resume --skip_preseq --reverse_stranded --skip_genebody_coverage --fcGroupFeaturesType "biotype"

I get the following:
ERROR ~ No such variable: gtf_makeSTARindex

[drpatelh]

Thanks @silviamorins ! I suspect it's something to do with the optional channel creations when specifying --gff as opposed to --gtf. Have you tracked down the bug? If not, I'll look into it 👍

[silviamorins]

@drpatelh sorry, no... I have looked into the code but it seems fine to me. I am not very into the channel structure, I guess that might be an issue

[silviamorins]

I just ran into this again, with another genome (apple). The content in here is parsed correctly, right? (i.e., is providing the information further?)

[drpatelh]

Sorry, this evaded my extensive TODO list. Apple genome! I didnt even know that existed 🥇

Ive had a look and cant see anything obvious with the relevant channels. May have to create a minimal test case that reproduce the error...

[lpantano]

I think this line:

rnaseq/main.nf

Line 427 in 37f260d

if(params.gff){

that converts GFF into GTF needs to be at the very beginning, so makeSTARindex use that GTF to make the index. Right now that makeSTARindex happens before the convertGFFtoGTF. (I think that is the issue)

[silviamorins]

Hi @lpantano , you're right, if the processes are executed in order that must be it. I will try to insert the change in my fork tomorrow and test it; if the error doesn't show up anymore I'll open a PR. Thanks!

[drpatelh]

This should be fixed now with:
#248