Skip to main content
Topic: In Silico predicted MS spectra - A good idea? (Read 10470 times) previous topic - next topic

In Silico predicted MS spectra - A good idea?

Dear,

I noticed that as of recently most of the compounds currently in the METLIN online database are accompanied by CID MS spectra in positive ionization mode. At first I was delighted because this information proofed to be extremely helpful for me identifying certain metabolites. However, my mood rapidly changed when I found out that most of these newly added spectra where in silico predicted.

My problem is that many of the in silico spectra are not indicated as such (e.g. Fig. 1), leaving me to think that they are based on actual empirical data (as was the case before). This caused me a great deal of confusion. Moreover, having seen many physical CID MS spectra, to me, all these predicted ones did not seem to have much to do with reality. I would like to know what program was used to make these spectra and how well this software preforms.

In conclusion, if you decide to add in silico spectra please indicate this very clearly and explicitly state the software used. But in all honesty I do not think they are very helpful and are going to cause a lot of confusion. I would like to hear some other opinions on this matter.

Regards,

Seb



Fig. 1) An in silico predicted spectrum which is not indicated as such.







Re: In Silico predicted MS spectra - A good idea?

Reply #1
I fully agree. I ran into the same problem recently.

If you click on the compound number the info is displayed on the spectrum, but not if you click on the View button. It took me an entire day before I realized that these spectra are CFM-ID spectra. Please put the "Insilico spectra" info at least on the "View" graphic if not on the result table.

Re: In Silico predicted MS spectra - A good idea?

Reply #2
Dear Meow,

Although I'm sorry to hear that you wasted time on this issue as well, I am glad  that I was not the only one. Maybe this could motivate the METLIN people to address this issue.

Furthermore I would like to reiterate my point about the low quality of the in silico spectra. In Fig 1 I show a low (5 V, bottom) and high (20 V, top) energy spectrum of  LPC(18:0) (a lysophosphatidylcholine) at m/z 524 as measured on our SYNAPT G2 ToF instrument. Fig2 are the low (bottom) and high energy (top) predicted spectra as found in the METLIN database. The differences are obvious.

I know for sure that 104 and 184 peaks in Fig 1 are characteristic for (L)PCs on any type of MS, however these are not obvious or completely missed by the predictions. Yet the predicted spectra show a whole bunch of peaks that are nowhere to be found in the actual spectra. This is just an example; I found more obviously wrong spectra. I really think these in silico specs should be removed from the database or at least should not be shown in the current form.

I would like to hear some opinions from the METLIN db curators if possible. However I am not sure how to contact them besides this forum.

Regards,

Seb



Fig 1) High (top) and low (bottom) energy spectra of LPC(18:0) as measure on a SYNAPT G2 ToF.





Fig 2) High (top) and low (bottom) energy spectra of LPC(18:0) in silico predicted as found on METLIN






Re: In Silico predicted MS spectra - A good idea?

Reply #3
Dear Seb,

First let me say that I'm sorry to hear that you are unhappy with our efforts. We have been doing a lot of developments with MS/MS data in the past few months and while not all of the spectra perfectly align, based on our initial review of the data with compounds that we did not have MS/MS data on, we feel that the in silico data does add value. 

To address your concerns that the spectra are not labeled clearly as "in silico" we will make the words “in silico” bigger so that it is now clearer. If you do not think that the in silico spectra are usable for your research in their current form please feel free to ignore them on the site. We have heard directly from many users that this new addition has been very helpful, and while not perfect they are happy to have this new information.

Finally let me thank Dr. Salek who contacted me via the METLIN contact page to alert me to this thread. I have attached an image of the where you can contact us from the METLIN website.

Best regards,

Paul 

PS Say hi to Oskar for me, if you see him soon.


[attachment deleted by admin]
~~
H. Paul Benton
Scripps Research Institute
If you have an error with XCMS Online please send me the JOBID and submit an error via the XCMS Online contact page

Re: In Silico predicted MS spectra - A good idea?

Reply #4
Dear Paul,

Thank you so much for your reply (and Reza for forwarding this to you). Clear annotation might indeed solve the problem.

In any case, could you provide a link to the software/literature to generate the spectra?

And I will say hi to Oscar (M?) :)

Regards,

Seb

Re: In Silico predicted MS spectra - A good idea?

Reply #5
Dear Seb,

We're planning a face change to METLIN in the near future with some added extra's. We'll see to it that there is also an option of removing the metabolites with insilico MS/MS data from the searches.
We haven't published anything on the MS/MS generation yet we have used a lot of the competitive fragment modeling ideas. 

Ah .. Sorry I thought I knew you from Dr. González - my Bilbao circle misinforms me. 

Best,

Paul
~~
H. Paul Benton
Scripps Research Institute
If you have an error with XCMS Online please send me the JOBID and submit an error via the XCMS Online contact page

Re: In Silico predicted MS spectra - A good idea?

Reply #6
Dear Paul,

OK, I'm looking forward to see the changes implemented. And the Bilbao circle is small and I will say hi to Oskar G when I see him :)

Cheers,

Seb

Re: In Silico predicted MS spectra - A good idea?

Reply #7
Hi Seb,

The software used to generate the in-silico spectra is CFM-ID - you can find more info here: http://cfmid.wishartlab.com/

Thanks for notifying the community and METLIN on this :-)

Cheers Justin

 

Re: In Silico predicted MS spectra - A good idea?

Reply #8
Hi Justin,

Thanks for the info.

S