MetFragCL 2.6.1: no individual MoNA scores

[rickhelmus]

Hi all,

I finally managed to have a look at the latest MetFrag CL version. However, from my CI test data I seem not not to get any MoNAScores>0 anymore.

This is what I was used to with 2.5.0:

                                SMILES                                                  compoundName individualMoNAScore
                                <char>                                                        <char>               <num>
1:            C1=CC2=C(C(=C1)O)N=CC=C2                                                 quinolin-8-ol             0.38374
2:            C1=CC=C2C(=C1)C(=CN2)C=O                                      1H-indole-3-carbaldehyde             0.87526
3:                    CCOP(=O)(OCC)OCC                                            triethyl phosphate             0.99489
4:          CCN(CC)C(=O)C1=CC=CC(=C1)C                                 N,N-diethyl-3-methylbenzamide             0.99290
5:  C1=CC=C(C=C1)N2C(=O)C(=C(C=N2)N)Cl                      5-amino-4-chloro-2-phenylpyridazin-3-one             0.40480
6: CN(C)CC(C1=CC=C(C=C1)OC)C2(CCCCC2)O 1-[2-(dimethylamino)-1-(4-methoxyphenyl)ethyl]cyclohexan-1-ol             0.53815

With 2.6.0 I get exactly the same results.

If if I use 2.6.1 and manually set the MoNA DB from here I actually get a few more:

                                 SMILES                                                  compoundName individualMoNAScore
                                 <char>                                                        <char>               <num>
 1:            C1=CC2=C(C(=C1)O)N=CC=C2                                                 quinolin-8-ol             0.47222
 2:            C1=CC=C2C(=C1)C=CC(=O)N2                                             1H-quinolin-2-one             0.99483
 3:        C1=CC=C2C(=C1)C=CC=[N+]2[O-]                                         1-oxidoquinolin-1-ium             0.88740
 4:              C1=CC=C2C(=C1)C=CNC2=O                                          2H-isoquinolin-1-one             0.99037
 5:            C1=CC=C2C(=C1)C(=O)C=CN2                                             1H-quinolin-4-one             0.62835
 6:            C1=CC2=C(C(=C1)O)N=CC=C2                                                 quinolin-8-ol             0.90691
 7:            C1=CC=C2C(=C1)C=CC(=O)N2                                             1H-quinolin-2-one             0.85572
 8:            C1=CC=C2C(=C1)C(=CN2)C=O                                      1H-indole-3-carbaldehyde             0.91169
 9:            C1=CC=C2C(=C1)C(=O)C=CN2                                             1H-quinolin-4-one             0.88355
10:            C1=CC2=C(C=CC(=C2)O)N=C1                                                 quinolin-6-ol             0.88252
11:              C1=CC=C2C(=C1)C=CNC2=O                                          2H-isoquinolin-1-one             0.12091
12:        C1=CC=C2C(=C1)C=CC=[N+]2[O-]                                         1-oxidoquinolin-1-ium             0.10941
13:                    CCOP(=O)(OCC)OCC                                            triethyl phosphate             0.99489
14:          CCN(CC)C(=O)C1=CC=CC(=C1)C                                 N,N-diethyl-3-methylbenzamide             0.99869
15:  C1=CC=C(C=C1)N2C(=O)C(=C(C=N2)N)Cl                      5-amino-4-chloro-2-phenylpyridazin-3-one             0.40519
16: CN(C)CC(C1=CC=C(C=C1)OC)C2(CCCCC2)O 1-[2-(dimethylamino)-1-(4-methoxyphenyl)ethyl]cyclohexan-1-ol             0.94936

Any clues?

Thanks,
Rick

[schymane]

I can confirm Rick's observation of zero similarity scores with the web version of MetFrag, this example is a spectrum from MassBank and should have a 1.000 exact similarity and currently has zero (I have just posted the settings via the Feedback form). @sneumann @meier-rene any ideas? Has it lost the link to the MoNA library somehow?

[meier-rene]

Hi,
I tried to reproduce the issue with the example you gave in the feedback mail. I can confirm the issue with the version on https://msbi.ipb-halle.de/MetFrag/. On the other hand starting the app locally works as expected. So there is hope that I can fix it. Hopefully tomorrow in the office I will find the issue.

[meier-rene]

The web versions are working fine now. There was an issue with an empty library folder in the deployment. Now I will look into the cli issue.

[meier-rene]

There was a issue with the loading of the reference library from the default resource folder, which only hits in case of running MetFrag from command line. Hopefully this is fixed. No need to provide the MoNA DB via file. This version is included by default. In the output you can see a log message about loading of 5 chunks of that library from default directory.
Please try version 2.6.3.

[rickhelmus]

Excellent! I can confirm things seem to work fine now :-)

Thanks,
Rick

[schymane]

Hi @meier-rene and @sneumann I'm reopening this because it looks like MetFragWeb (and Beta) have both reverted to 2.6.1 and part way through teaching we have zero MoNA scores again - I also just submitted feedback with the default example. Here's what the version stamp looks like ...

[meier-rene]

I can confirm, but its something different. I could quick fix it by restarting the software at msbi.ipb-halle.de/MetFrag. I need to follow the logs until I find something suspicious. I also did not restart /MetfragBeta. Maybe I can find the reason there.
With regards to the version number: maybe I forgot to update it. Dont know where. 🙈

[schymane]

OK yes thanks @meier-rene - I can confirm the MS/MS values look much better now on the main website (curiously, it was not zero completely across the board, there were some very poor scores coming through occasionally, but for compounds that should have very high scores - and we got zero scores for many compounds we know are in the library and should have scores, like the test example). Maybe @sneumann knows where to update the version number, I thought with that fancy hash that it was automated ;-)

[schymane]

Hi @meier-rene @sneumann - so, it seems it's more complicated ... the example data is still returning a nice match value of 0.9 (as in the screenshot above), but we still are not getting good (or any) match values for most of the basic amino acids that should be in MassBank (even with really nice spectra), and I just tried an example of caffeine directly from MassBank and that also returns a match value of zero... (I also just sent the settings via the feedback channel).
So, it looks like we have at best a very incomplete spectral library connection - or something very strange is going on ...

https://massbank.eu/MassBank/RecordDisplay?id=MSBNK-LCSB-LU119903&dsn=LCSB

[meier-rene]

I followed the example you send by email and I can reproduce your findings. I have no clue at the moment and digging through the code. It works fine in command line mode but something is strange running in the webapp.