Metabolomics data analysis tools

[lamccue]

Discussion of underlying core software used for LC-MS metabolomics

NMDC uses CoreMS

• Fully open-source: CoreMS
• Sponsored, funded and maintained by a DOE-BER User Facility
• Database for compound identification is open-source: Metabref
  • Currently has metabolites for GC-MS and lipids for LC-MS data
• CoreMS is a comprehensive framework that:
  • works with a variety of raw data input formats and vendors (.mzml, .cdf, .hdf5, .raw, etc) allowing end to end data processing
  • includes modules for all of the key processing steps for raw data: signal processing, peak picking for m/z and retention time, centroid detection, MS1 deconvolution of mass features, molecular formula searching on MS1 or MS2 spectra, spectral search using entropy similarity and additional metrics
  • provides molecular formula for the identified peaks, regardless of whether there was a match to a compound in the database

GNPS

• GNPS provides workflows that are kind-of open-source? (Unclear about this...and you maybe need a token to use workflows on the web interface?): UCSD workflows
• Database for compound identification appears to be crowd-sourced and not available except through the GUI-based workflows (i.e., database is not downloadable)
• Workflows support input of raw data in one format (.mzml), then use third-party software for the signal processing steps for that data
  • Signal processing uses MZMine2
  • Centroiding (peak profile to centroid data conversion) relies on MSConvert
• Workflows do not resolve molecular formula, which are especially useful for checking annotations or providing some annotation in the case of a no-database-match scenario
• Documentation suggests that the workflows are in GUI format, and do not appear to be CLI compliant; thus their "customizable" workflow in the GUI will not be reproducible
• The molecular networking methods unique to GNPS are designed for a further downstream analysis (beyond the primary data processing), by providing cross-sample and cross-mass-spec experiment exploration

Common LC-MS metabolomics issues

• Annotations of compounds are the bottleneck for any LC-MS metabolomics analysis.
  • The annotations are critical to connect date across omics types.
  • Currently, everyone in the community relies on sifting through the suggested annotations that a software tool provides, which is a manual process requiring mass spec expertise
  • We need a highly curated database that we trust – the entire metabolomics community needs this, but there are no good, public options.

@emileyfadrosh @aclum @corilo @kheal @mslarae13

[aclum]

It is my understanding that the workflows are available in nextflow ex https://github.com/Wang-Bioinformatics-Lab/Feature_Based_Molecular_Networking_Workflow/blob/master/nf_workflow.nf

[aclum]

If the workflow above is correct the specta db would be https://github.com/Wang-Bioinformatics-Lab/Feature_Based_Molecular_Networking_Workflow/tree/master/data/mzmine2/gnps_featurefinding/spectra

[kheal]

@aclum that spectra db is what is being searched not annotated reference spectra to compare to. The "data/libraries" would be the reference input but that is not supplied.

[kheal]

That workflow doesn't seem to include any of the signal processing - its inputs are already data that have been somewhat processed (likely by two other tools, MSConvert and MZMine2).

[aclum]

is this more what you were looking for https://gnps-external.ucsd.edu/gnpslibrary?

[kheal]

Ah, that is a very helpful link @aclum thank you. It demonstrates the root of the challenge for this workflow perfectly, which database do we use, how do we know how much trust to put into each, and is there sufficient and consistent metadata regarding these reference spectra and their associated molecules to be able to provide the confidence in our annotations according to best practices? I don't have answers for that, but this is a good reference once we have capacity to assess.

[aclum]

These are tasks that got run in GNPS2 for bioscales by Ming as part of the subcontract
https://gnps2.org/status?task=55e70850bcfe4b47b9fed6f9d229a2eb (feature_based_molecular_networking_workflow).. unclear still what library as used as this appears to be set as an env variable (LIBRARYLOCATION/LC/LIBRARY)
https://gnps2.org/status?task=0a63ad9f5f3e44c88891ca7a1855c5e7 (classical_networking_workflow) input is mzML data. This is the repo that has the nextflow workflow for classical_networking_workflow

[aclum]

Documentation for the feature based molecular networking workflow https://wang-bioinformatics-lab.github.io/GNPS2_Documentation/fbmn/
classical molecular networking https://wang-bioinformatics-lab.github.io/GNPS2_Documentation/classical/

[aclum]

example input libraries: https://gnps2.org/filebrowser/files/nf_tasks/55e70850bcfe4b47b9fed6f9d229a2eb/input_libraries/