MSP as export format for MassBank-data

[sneumann]

Hi, in addition to #31 and #32 we can think about *.msp
as export format for the releases. @meier-rene has code
up the sleeve that exports MB records to some flavour of *.msp files.
Yours, Steffen

[tsufz]

Really appreciated!

[schymane]

Somewhat related ... @adelenelai shared this earlier today
https://www.researchsquare.com/article/rs-44215/v1

They do not yet seem to support MSP (on a superficial look) but maybe an extension?

[tsufz]

Jep, They are part of NFDI4chem.

[meowcat]

I usually use http://prime.psc.riken.jp/compms/others/main.html#Massbank2msp (usually followed by the LIB2Nist to convert it to a NIST library). This works quite well with the appropriate settings. But possibly @meier-rene has something better?

[sneumann]

IIRC Massbank2msp is a windows program, so it would not work nicely in a continuous integration pipeline. Correct ?

[tsufz]

Github should provide a Windows based runner... However, it might be a nicer way to write an own parser. The format is super simple and this should be doable in some hours:

NAME: Mellein; LC-ESI-ITFT; MS2; CE
PRECURSORMZ: 179.0697
PRECURSORTYPE: [M+H]+
INSTRUMENTTYPE: LC-ESI-ITFT
INSTRUMENT: Q-Exactive Orbitrap Thermo Scientific
Authors: Justin B. Renaud, Mark W. Sumarah, Agriculture and Agri-Food Canada
License: CC BY-SA
SMILES: CC1CC2=C(C(=CC=C2)O)C(=O)O1
INCHI: InChI=1S/C10H10O3/c1-6-5-7-3-2-4-8(11)9(7)10(12)13-6/h2-4,6,11H,5H2,1H3
COLLISIONENERGY: 10(NCE)
FORMULA: C10H10O3
RETENTIONTIME: 3.44
IONMODE: Positive
MASSBANKACCESSION: AC000001
Links: INCHIKEY KWILGNNWGSNMPA-UHFFFAOYSA-N; CAS 17397-85-2; PUBCHEM CID; CHEMSPIDER 26529; KNAPSACK C00000550; COMPTOX DTXSID60891794;
Comment: PrecursorMz=179.0697, PrecursorType=[M+H]+, InstrumentType=LC-ESI-ITFT, CE=10(NCE)
Num Peaks: 5
133.0648 21905.33203125
151.0754 9239.8974609375
155.9743 10980.8896484375
161.0597 96508.4375
179.0703 72563.875

[tsufz]

The tool of Riken is obviously not running in command line mode and thus no chance for automation. Hence, an offline workaround and using a runner mode is not possible. :-(

[meowcat]

At one point I started (with @michaelwitting and @Treutler) work on a converter that uses schema templates, so it would read/write arbitrary formats by just adding new templates. I still find the idea neat but we never really materialized the thing. A simple MSP writer is easier to do, of course.
https://github.com/meowcat/MSnio

[sneumann]

Another initiative (which can still use the schema template idea) is
https://github.com/rformassspectrometry/Spectra/ which can then
use different MsBackEnds https://github.com/sneumann/MsBackendMsp

Also note that we have started to document parts of the MassBank environment
in https://github.com/MassBank/MassBank-documentation/ so they are available
via the MassBank web frontend: https://msbi.ipb-halle.de/MassBank/
Then it is possible to have updated documentation without re-deploying the web app :-)

Yours, Steffen

[tsufz]

Really appreciated :-)

Best,
Tobias

[michaelwitting]

@meowcat Indeed, this is something we should follow up. Maybe the Spectra package is a good opportunity to revive it. I really like the backends and the Spectra objects. It is very easy to construct. I have written a backend for MassBank records: https://github.com/michaelwitting/MsBackendMassbank. If I find time soonish I will implement a write for MassBank records.
The MSnio package could be used to convert the namespace from one backend to another.

[tsufz]

Nice, but we need a Java backend to integrate somewhere in CI, for example in the validator. We parse the records for the checks anyway. Some additional parsing and export to the *.msp and some downstream processing to compile meaningful *.msp collections might be possible. @meier-rene, what do you think?

[sneumann]

@meier-rene has now worked on an export for MSP that will appear in our releases. Yours, Steffen

[tsufz]

By today, this appeared! Many thanks to @meier-rene for your efforts!

https://github.com/MassBank/MassBank-data/releases/tag/2020.11

Time to close this issue.

[tsufz]

Reopen... It is a pity that the original record number is not included and we should give some credits to the authors. Could be parsed to the comment field.

[meier-rene]

I will include this. Thanks for the comment. This converter to MSP is pretty new, so I expect some more issues. Please don't hesitate to report them.

[meowcat]

Yep. First, thanks a ton! Second, the original accession ID would be really great.

Then finally, not essential but worth a thought, the MSP format used by NIST (which I consider "the reference") uses a mixture of CamelCase (PrecursorMZ) and Capitalized_snake_case (Ion_mode). This is from an obfuscated version of an original NIST record:

Name: Obfuscamycin-methyl
Notes: Consensus spectrum; Nreps=9/9; Mz_diff=-0.2ppm; Vial_ID=3432; Metabolite_2016_01_29_ID=65432; micromol/L in water/acetonitrile/formic acid (50/50/0.1)
Ion_mode: P
Instrument: Thermo Finnigan Elite Orbitrap
Instrument_type: HCD
Ionization: ESI
Collision_energy: NCE=15% 11eV
Collision_gas: N2
Sample_inlet: direct flow injection
Spectrum_type: MS2
Precursor_type: [M+H]+
PrecursorMZ: 433.0658
InChIKey: WTZUASDHJKHKFA-UHFFFAOYSA-N
Synon: (The first synon could/should be the IUPAC name)
Formula: C54H13ClF3NO4
MW: 432
ExactMass: 432.058521
CASNO: 12345
NISTNO: 34567
ID: 8765432
Comment: NIST Mass Spectrometry Data Center
Num peaks: 3
123.222 654.12 "C13ClF3H10NO2=p-C3H1O2/0.1ppm;C34ClH11NO4=p-H3F3/-7.6ppm 8/8"
341.0294 77.40 "C11ClF3H10NO3=p-CH4O/-0.5ppm 8/8"
398.0558 999.00 "p/0.0ppm 8/8"

In the further future, we could also include the peak annotations.

We once compiled correspondence between record format fields here, maybe that helps:
https://docs.google.com/spreadsheets/d/1IGgtA6FL5lt9ajDSb0WR90qQfHtebDtcfyjX86tC_JQ/edit#gid=0

[meier-rene]

@tsufz Your issue is solved. I uploaded a RIKEN msp version which includes a DB id.

@meowcat The first msp I uploaded was not in the NIST format, but the RIKEN format. I've chosen this because its more used in our place, but I don't want to argue which one is the reference ore more useful. Our converter can also create NIST dialect of msp and thats why I also uploaded a NIST format msp. Future releases will include both version, but I will not implement a third version. 😄 Your record format cheat sheet will become useful in the future. Thank you for sharing! I would appreciate it, if it would stay accessible.