Qemistree-Sirius stalls out a 'predict-fingerprints' step

130 views
Skip to first unread message

Brian Sedio

unread,
Apr 12, 2021, 10:03:09 AM4/12/21
to GNPS Discussion Forum and Bug Reports
Hi Priya, Everyone,

Due to the server issues with Qemistree on GNPS, I'm running it through qiime on my laptop. Everything works fine until I get to the 'predict-fingerprints' step. Here are examples using the demo data:

The following 'comput-fragmentation trees' and 'rerank-molecular-formulas' work fine, and generate output that looks good to me:

qiime qemistree compute-fragmentation-trees --p-sirius-path '/Users/briansedio/Documents/Sirius/sirius-osx64-4.0.1/bin' \
  --i-features sirius.mgf.qza \
  --p-ppm-max 15 \
  --p-profile orbitrap \
  --p-ionization-mode positive \
  --p-java-flags "-Djava.io.tmpdir=/Users/briansedio/Documents/QemUT/demo-qemistree/javatemp -Xms16G -Xmx64G" \
  --o-fragmentation-trees fragmentation_trees.qza


qiime qemistree rerank-molecular-formulas --p-sirius-path '/Users/briansedio/Documents/Sirius/sirius-osx64-4.0.1/bin' \
  --i-features sirius.mgf.qza \
  --i-fragmentation-trees fragmentation_trees.qza \
  --p-zodiac-threshold 0.95 \
  --p-java-flags "-Djava.io.tmpdir=/Users/briansedio/Documents/QemUT/demo-qemistree/javatemp -Xms16G -Xmx64G" \
  --o-molecular-formulas molecular_formulas.qza

Then 'predict-fingerprints' seems to very briefly start python and java, then sit idle for days, even with the demo data:

qiime qemistree predict-fingerprints --p-sirius-path '/Users/briansedio/Documents/Sirius/sirius-osx64-4.0.1/bin' \
  --i-molecular-formulas molecular_formulas.qza \
  --p-ppm-max 20 \
  --p-java-flags "-Djava.io.tmpdir=/Users/briansedio/Documents/QemUT/demo-qemistree/javatemp -Xms16G -Xmx64G" \
  --o-predicted-fingerprints fingerprints.qza

I didn't have this problem a few months ago. Do you think this is an issue with Sirius? Do the preceding steps not need to use the Sirius server, and the reason 'predict-fingerprints' stalls out it because of a server issue? 

Thank you very much for your help!
Brian

Allegra Aron

unread,
Apr 12, 2021, 5:56:33 PM4/12/21
to GNPS Discussion Forum and Bug Reports
Hi Priya, Ming, Brian, and all,

I also had trouble running Qemistree jobs on GNPS last Friday. I tried to run 3 or 4 jobs, and I kept getting the following error message: "Workflow node [qemistree] was forcibly terminated by the computational back end. The most likely reason is for exceeding this node's memory usage quota." I thought perhaps, I had too many features, but the fact this failed on multiple jobs plus your message (Brian) indicated to me that something is wrong with the workflow. If this is the case, I can run in qiime... just was hoping to figure out whats up!

Thanks,
Allegra

Brian Sedio

unread,
Apr 12, 2021, 9:54:16 PM4/12/21
to GNPS Discussion Forum and Bug Reports
Hi Allegra,

Have you tried going through each step using Qiime2 locally, like I described, above? Are you running into any problems with that workflow?

Cheers,
Brian

Allegra Aron

unread,
Apr 12, 2021, 9:56:56 PM4/12/21
to Brian Sedio, GNPS Discussion Forum and Bug Reports
Hi Brian,
I haven't done that yet because the Qemistree workflow in GNPS worked just the week before last, so I was wondering if there was a bug ... but I'm happy to go through the processing in Qiime2 to see how that goes.
Best,
Allegra

--
You received this message because you are subscribed to the Google Groups "GNPS Discussion Forum and Bug Reports" group.
To unsubscribe from this group and stop receiving emails from it, send an email to molecular_networking_b...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/molecular_networking_bug_reports/0de0ddd9-679c-4297-ab41-18ddd6d55043n%40googlegroups.com.

Brian Sedio

unread,
Apr 13, 2021, 11:23:34 AM4/13/21
to GNPS Discussion Forum and Bug Reports
Hi Allegra,

Ok, let me know if you are able to follow through the Qemistree demo using the demo data on the Github page (https://github.com/biocore/q2-qemistree).

The demo uses Sirius 4.0.1, but Sirius is now on version 4.7.4. The latest version is structured differently, and I can't figure out how to modify this line to make it appropriate for the new version: --p-sirius-path '/Users/briansedio/Documents/Sirius/sirius-osx64-4.0.1/bin'

So please do let me know if you find success with either version using the demo data, so we can figure out if the issue is with Sirius or not.

Thanks!
Brian

Allegra Aron

unread,
Apr 13, 2021, 12:12:31 PM4/13/21
to Brian Sedio, GNPS Discussion Forum and Bug Reports
Thanks Brian! I'll let you know how it goes ....

Best,
Allegra

You received this message because you are subscribed to a topic in the Google Groups "GNPS Discussion Forum and Bug Reports" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/molecular_networking_bug_reports/b6PvJvEA2GQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to molecular_networking_b...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/molecular_networking_bug_reports/c2ceeb9a-6e53-4f04-af7c-5056c900f6a5n%40googlegroups.com.

Anupriya Tripathi

unread,
Apr 14, 2021, 5:45:36 PM4/14/21
to Allegra Aron, Brian Sedio, GNPS Discussion Forum and Bug Reports
Hi Brian, Allegra, 

If your jobs are stalling at the `predict-fingerprint` stage, this means that the CSI: FingerID server is unresponsive. I'd do a few things to follow-up: 1) run predict fingerprint with a single file - does it still stall? 2) kill the job and restart - this step is usually quicker so it's unusual that it's stuck for days, 3) if the problem persists, post an issue on the Sirius GitHub page - this might be the fastest way to resolve as the Boecker lab maintains this server. 
Also please double-check that you have a stable internet connection throughout the process!

Please let us know how it goes.

Thanks, Priya

Anupriya Tripathi, PhD


Brian Sedio

unread,
May 25, 2021, 11:51:45 PM5/25/21
to GNPS Discussion Forum and Bug Reports
Hi Priya and Allegra,

It looks like the 'end of life' of Sirius 4.0.1 on 04/30/21 was responsible for all of this: https://bio.informatik.uni-jena.de/2021/04/sirius-4-0-1-end-of-life/

Do you know if there are plans to reconnect GNPS (and Qemistree in particular) with the latest Sirius version? It would be super powerful to take advantage of the CANOPUS as well.

Thanks again!
Brian

Reply all
Reply to author
Forward
0 new messages