secondary analysis

[splaisan]

Dear Vladimir,

I am finally starting to use your pipeline and fixed a few issues at my side that were crashing step2.
I now have al expected folders but do not find how to proceed and get classification (I now ran a manual blast) as well nice bar plots and also phylogeny plots.

I expect this can be done using Qiime2 but I am still not too familiar with it and which files need be imported to create the right objects for plotting.

Would you by any chance have a tutorial explaining the next steps for visualisation to obtain results similar to the pb-16S-nf pipeline does for 16S. Ideally, an extension to your current https://next-its.github.io/ doc would be perfect with results from some test data.

Thanks in advance

[vmikk]

Hello Stephane,

I'm sorry for the late reply.

Currently, NextITS does not perform taxonomy annotation of the resulting sequences. In our group, we primarily use BLASTn for this purpose (as it is the most sensitive classifier among all we have tested) in conjunction with the EUKARYOME database. However, BLAST is extremely slow and should be run in parallel on separate nodes if you have a large number of sequences to annotate. Additionally, the resulting table requires curation, which is not yet automated.

In general, since the output OTUs are in FASTA format, any taxonomy annotation tool should support it. We might consider adding a taxonomy annotation step, but we need to decide on the most viable approach.

Regarding the test data and documentation, updating them is certainly a good idea, and it is already on our to-do list.

With kind regards,
Vladimir

P.S. We have added support for asymmetric barcodes to NextITS.

[splaisan]

Thanks for this answer.

I could classify using QIIME2 and it worked quite well, I attach my script here in case it can help you or any other user.

I did not use your dual-barcode support yet because my data gets demuxed during the processing of the kinnex raw data and I can submit the ITS samples to your pipeline as a fastq folder while analyzing the 16S fastq data with 16S-HiFi-nf.

Best regards,
Stephane

[splaisan]

Dear @vmikk,
I finished the analysis of a large number of samples and am puzzled by the fact that the wide table only report 1 or few different OTUs with very low numbers while the read data can be as large as 50000 PacBio ITS1-5s-ITS2.

As I canot test this with a validated input I am not sure it is true or wether the pipeline deduplicated OTU counts to 1 by mistake.
Can you please confirm that the wide (and long) table reports the true count of reads that match each OTU in each sample after removing artefacts?
If you know of one, could you please share a SRA link with validated mock data so that I try your pipelne against a known OTU frequency table. As long as I cannot trust the results I will not use this pipeline in production.
Sorry to insist but I quite urgently need a valid ITS pipeline for sequencing data to come.

would SRR30980783+SRR30980784 be a good pick and could you please share the corresponding step2 wide table for it?

[vmikk]

Hello Stephane,

that the wide table only report 1 or few different OTUs with very low numbers while the read data can be as large as 50000 PacBio ITS1-5s-ITS2

That's strange. Is it the same for the pre-LULU data?
There should be logfiles for each process in the working directory (.command.out files), let's check the dereplication and clustering steps.

Can you please confirm that the wide (and long) table reports the true count of reads that match each OTU in each sample after removing artefacts?

Yes, these tables should contain read counts for the processed data.

would SRR30980783+SRR30980784 be a good pick

These should work well. The amplicon is rather large as LR7Angs is located near D5 of the LSU, but I believe that ITSx should recognize it well.

Unfortunately, I can not test these data myself now as I'm traveling.

[splaisan]

I am running now the two SRA samples...

Considering my bad analysis results, when I look in Step1/08_RunSummary in the excel file I see very few ITSx classified counts for a large number of filtered reads.

I add here a picture of the top of this table.

Does this mean that ITSx failed to find more OTUs and thereby lead to a very small count?
I attach both nextflow logs: .1 being Step1 and the other step2. Can you please tell me which log files should I look at to find a cause for the issue?

nextflow_logs.zip

[vmikk]

From the table, I see that the amplicon looks good in general (almost no "Primer artefacts"), but ITSx definitely failed.
Which primers are you using? It might be that one of them is located in the HMM-recognition sites (e.g., the exact border of SSU and ITS1), and hence ITSx is not working.
If this is the case, you may skip ITSx and just trim the primers with --its_region none.

[splaisan]

You are probably spot on.
The primers used were:
ITS1ngs TCCGTAGGTGAACCTGC instead of TACACACCGCCCGTCG (apparently just at the end of SSU or even walking onto ITS1 as you suggested)
ITS4ngs TCCTSCGCTTATTGATATGC instead of CCTSCSCTTANTDATATGC (identical except first base with no effect)
I start a new analysis with your edit to see what it gives.
Thanks for your kind feedback.

[splaisan]

Seems to have worked on a few samples, I get a very diverse list of counts as expected.
I am now running the big experiment once more.
Great help, thank you