MS2Rescore config change to prevent crash on 16GB computer

[rohitsuresh1407-cpu]

Hello,

I am running DDA analysis from data generated on the timsTOF Bruker instrument. I am currently using a 16GB ram machine on azure with 100GB storage. I was able to change configurations to run comet with this machine (by restricting CPU threads to 2 and memory to 7GB, while also editing core parameters restricting numerical threads to 2 and spectrum batch size to 200). While the comet ran successfully, it kept crashing at the MS2Rescore step (error message below). I was wondering what parameter to fix to make MS2Rescore step run.
Thanks in advance!

Basic details:
nf-core/mhcquant: 2.x
Nextflow: 25.04.7
Size: Standard E2s v3 (2 vcpus, 16 GiB memory)
Apptainer 1.3.1
OS: Ubuntu 22.04

Nextflow code:
nextflow run ~/mhcquant/main.nf
-profile singularity
-c ~/mhcquant_run/conf/xyz.config
--input sample_sheet.tsv
--fasta HUMAN_2014_concat.fasta
--skip_decoy_generation true
--ms2pip_model timsTOF
--ms2pip_model_dir /home/xyz/ms2pip_models
--outdir results_main

I have attached my error message in the pdf below.

Nextflow error.pdf

[jonasscheid]

Hi!
In your case of very limited compute ressources, I would suggest changing process labels of compute intensive tasks (such as comet or ms2rescore) via the withLabel configuration file. See here https://www.nextflow.io/docs/latest/config.html#process-selectors

In nf-core pipelines these compute heavy tasks are labeled with process_high. There are no MS2Rescore settings to predict in batches unfortunately, so you might need to find out how much RAM you need for your data. timsTOF data is notriously big, so I would give process_high your max mem -1GB or so. Hope that helps!

[rohitsuresh1407-cpu]

Hi, thank you for your response!
I successfully got through the ms2rescore step, after making the changes you suggested, but it crashed at the percolator step. the error message attached to the pdf below. I have attached my full conf also below for your reference.

Also on a different note, I am processing DDA data, and I am interested in extracting and visualising the chromatograms using skyline. The output files on mhcquant2 are in the form of ms2rescore.idXML, but skyline requires pep.XML or mzid files. I tried using IDconvert from openMS, but some data and spectra mismatch occurs due to that. Do you have any suggestions on how I could move ahead in this?

Error document.pdf

Thanks you!

[jonasscheid]

Your welcome!

Your error is
Process requirement exceeds available CPUs -- req: 6; avail: 2

If you look at the percolator specifications you see it has process_medium, which requires 6 cpus. You can again change the settings of this label to only use 2 cpus

I am interested in extracting and visualising the chromatograms using skyline.

There is afaik no proper way in doing this. You have two options afaik:

• Use annotate_ions flag to get two additional output tsv files that contain all information you need to plot any peptide spectrum in the FDR-filtered list by your own
• Use spectrum_utils to plot specific peptides with their respective spectrum_reference (by searching this information in the pout_filtered.idXML)

We are currently working on an OpenMS Webapp that takes mzML + any ID file format (including idXML) to streamline spectrum visualization. I can let you know once its out.

[rohitsuresh1407-cpu]

Thanks for your reply!
It would be amazing if you let me know about the Webapp. I will revert back on the 16GB as well. Right now while using a different VM to run my pipeline, I had another small doubt. As per the MHCquant2 paper, the comet parameters that I wanted to change, especially precursor charges, would it be possible to edit it under params in config file if so in what format, as it does not seem to accept the changes. I was able to change from trypsin to unspecific, but changes that I wanted to incorporated as mentioned in the pdf document were not exactly implemented. May I know what has to be done.
Thank you!

Error document 2.pdf

[jonasscheid]

Please see the Warning section in the Usage section: https://nf-co.re/mhcquant/3.0.0/#usage

Please provide pipeline parameters via the CLI or Nextflow -params-file option. Custom config files including those provided by the -c Nextflow option can be used to provide any configuration except for parameters; see docs.

So either you use the -params-file flag with a file containing your params map, or you just specify them via the CLI

[rohitsuresh1407-cpu]

Thank you for your response and apologies for my delayed reply!

Since I am doing immunopeptidomics on HLA1 peptides, I have been successful in changing most parameters as per the recent MHCquant2 paper. What I seem to struggle is changing peptide length (8-14 that I want) from the default (6-54). I was wondering if these are editable parameters? or if I have to change them in the main.nf file? Could you walk me through how you made these comet parameters edits when you ran the MHCquant2 pipeline?
I tried both ways you mentioned above, but the peptide length keeps evading my edits.
Thank you!

[jonasscheid]

You do not need to change anything for comet. Use the post-percolator filter peptide_min_length and peptide_max_length CLI flag to do as in MHCquant2 paper

[rohitsuresh1407-cpu]

Ah okay thanks a lot!
I also noticed another discrepancy. When I was figuring out a mismatch, I noticed that while my pout.idXML and ms2rescore.idXML have around 300k peptides, only 77k seemed to have retained the PSM reference spectra. This means I cannot match the remaining back to their original mzml. Any reason for this?

[jonasscheid]

only 77k seemed to have retained the PSM reference spectra.

Depends on your fdr-level. Default is peptide-level-fdr, which keeps the best spectrum per peptide and sample_condition group in your samplesheet see here

This means I cannot match the remaining back to their original mzml. Any reason for this?

You can. For each PeptideIdentification in the pout_filtered (also non-filtered). There is a merge index <UserParam type="int" name="id_merge_index" value="2"/> , which corresponds to to the index of the spectra_data list e.g. <UserParam type="stringList" name="spectra_data" value="[file1.mzML, file2.mzML,file3.mzML]"/>
Its not pretty, agreed. It would be great to have this information in the output file as well. I think it might be nice to have an USI in the output tsv. I opened an issue in OpenMS OpenMS/OpenMS#8470