Improve pipeline tests for different protocols

[apeltzer]

Description of feature

Should be done in two steps:

• Finding proper testdata that is published somewhere
• Integrating these into testData
• Using the new config profile(s) as in Move protocols to config profile(s) #364 and then testing these properly

[lpantano]

this could be an option, can you take a look and see if we can use some of these samples:

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149513

and script confirming the trimming:

https://github.com/besarka16/Benchmarking-of-small-RNA-seq/blob/master/scripts/1_trimming.sh

[apeltzer]

Looks like we can use these to me - lets see how easy it would be to add these to nf-core/testdatasets and then run them using the profiles. Shouldn't be too hard and we can get some proper tests then running.

[lpantano]

that would be great!

[atrigila]

Hi @lpantano, we are having some issues in CI testing due to, what I think is, making several requests from mirBase. Is it possible to use any other source for the hairpin.fa and the mature.fa? Ideally, we would host the data in the nf-core test-dataset repository if the licence allows for it. For example, could we use RNAcentral? Is it possible to obtain equivalent datasets from this source?
Thank you!

cc @nschcolnicov

[lpantano]

as far as I know, tools need those file from miRBase, we could potentially put them nf-core test-dataset, I don't think the license is a problem. I don't see anything in their webpage that will limit that. I think the hairpin.fa is the only one in RNAcentral, but as well we will need the GFF file from miRBase and the mature.

[apeltzer]

If the licence is no problem, we can certainly dump this on S3 and use that URl.

[atrigila]

@nschcolnicov Nico wrote an E-mail to confirm license

[nschcolnicov]

Yes! I sent an email to '[email protected]' last week, but I'm not sure when I will get an answer, I'll update this ticket as soon as I have news

[nschcolnicov]

The protocol parameter is no longer used, and protocols are now loaded via the -profile argument. We are already testing illumina and nextflex protocols via the nf-tests, and the only difference across all protocols is just allocating different values to the same three parameters: clip_r1, three_prime_clip_r1, three_prime_adapter.
There is no need to have a dedicated test also for qiaseq and cats protocol, since we are already validating the workflow that is impacted by them via the illumina and nextflex protocols.
So, we will remove the deprecated "protocol" parameter from the test profiles, and close this ticket

[nschcolnicov]

PR was merged #456