nf-core rnaseq pipeline - determination of the correct strandedness for RNASeq libraries

[pnatarajan9]

Hello,
I have some questions regarding strandedness determination for rnaseq libraries:

1. With RSeQC using actual alignments to genome to determine the strandedness of libraries (compared to salmon method that uses pseudo mapping method), would you say that we should be using the RSeQC determined strandedness for the downstream salmon quantification?
2. If the answer to the above Q is "yes", can you make an option available to USE the RSeQC determined strandedness for salmon quantification? Currently the pipeline seems to be wired to use the strandedness as determined by salmon method, but simply points out if there are any discrepancies (among choice made in samplesheet.csv, salmon and RSeQC determined strandedness).
3. Here is a situation we have with nearly half of the samples in a project where there is a disagreement in the strandedness determination outcome. The details are presented here:
   The command used for running the pipeline -
   nextflow run /software/nfcore/pipelines/rnaseq/3.12.0/workflow --input ./samplesheet.csv --outdir results -profile singularity -config /nfcorepipelines/research.config -config /nfcorepipelines/big_dataset_workflow_resource.config --genome human-ensembl-grch38-r98 --skip_bigwig --remove_ribo_rna false

For 8 out of the 18 samples analyzed, salmon has assigned "un-stranded" (as shown in the attached image) and the rest of the 10 samples had agreement between salmon method and RSeQC method (with "reverse" for strandedness assigned). We had set "auto" for strandedness for all the 18 samples in our samplesheet.csv file.

What are your recommendations in general?
Do we need to re-run the pipeline when there are discrepancies (between salmon vs rseqc methods) and which one does one go with? I am sure users would have asked these Qs already, but it helps to be able to have an option to choose the recommended method for downstream salmon quant step.

Thank you!
Padma

[pinin4fjords]

I would recommend you try the latest pipeline version. The latest 'duplicates' the rows so you get a clearer view of what Salmon says relative to RSeQC- see the example in the docs: https://nf-co.re/rnaseq/3.18.0/docs/output#infer-experiment. I also did quite a bit of work to harmonise how the strandedness is inferred by the two methods. I think you may find this issue just goes away with the latest version.

I could talk some more about what it might mean if they're still very at odds (see a related Slack thread), but please try the lastest version first.

[pinin4fjords]

Closing after a month of inactivity after my last response.

[pnatarajan9]

The latest version of rnaseq (3.18.0) worked for us as per your suggestion. Thanks, Padma

…

On Tue, May 27, 2025 at 6:30 AM Jonathan Manning ***@***.***> wrote: *pinin4fjords* left a comment (nf-core/rnaseq#1539) <#1539 (comment)> Closing after a month of inactivity after my last response. — Reply to this email directly, view it on GitHub <#1539 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEHIZ25HYASD3XEP7W4FVKD3ARSINAVCNFSM6AAAAAB3D5XSV2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMJSGUZDMOJUHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>