Skip to main content
Topic: MS-DIAL and MS-FINDER library limits (Read 720 times) previous topic - next topic

MS-DIAL and MS-FINDER library limits

Hello,

I have converted the 2020 NIST HR-MSMS library to .msp format and tried using it in MS-DIAL and MS-FINDER. I had to amend some keys as the NIST ones didn't match that of the MS-DIAL libraries. Once I had done this small subsets of the entire NIST library worked. However the whole library (3Gb) doesn't provide any annotations in MS-DIAL or MS-FINDER. Is there a limit to the size of a .msp file these programs can use? Is there some way that I can test to see if MS-DIAL or MS-FINDER are loading the library correctly?

Thanks in advance.
Chris

Re: MS-DIAL and MS-FINDER library limits

Reply #1
Hi Chris,

At least size is not a problem. I am using concatenated libraries in .msp formats that go well above 5-6 GB and no problems at all!
Most likely, its the "keys/ fields" with formatting issues.

For a test run, I always take  few known spectra from this 3GB library and create a small .msp file to see if it works with say 0% or 5% cosine similarity cut off to make sure at least the library is recognized!

Hope it helps.

Thanks,
Biswa

Re: MS-DIAL and MS-FINDER library limits

Reply #2
Hi Biswapriya,

thanks for your response. Good to know that there's no limits to this brilliant software! I've run the same tests as you and it seems to work fine. I've also imported a subset of the subset as query spectra into MS-FINDER as a sanity check and it does a pretty good job matching them up. Not perfect though, which is worrying. For example, searching a series of query spectra for PC(16:0, 18:2) across varying collision energies I get five spectral matches returned from the library for query CE of 14, none for CE16 and seven for CE20. Visually inspecting the CE16 spectra there is little difference between that and adjacent CE spectra, as you would expect. So I don't understand why I get no matches. It would be great to have a variety of matching algorithms, as are available in MS-DIAL (dot product, reverse dot product, unweighted dot product, presence peak score, total MS/MS similarity score). I can't see where to vary the match factor threshold in MS-FINDER either and I can't tell what the algorithm its using is.

Do you use your own scheme for .msp keys? I am thinking I will edit the libraries to remove fields that the software doesn't use, in case those are the problem.
Cheers,
 Chris

Re: MS-DIAL and MS-FINDER library limits

Reply #3
Further to previous, I have duplicated the inconsistent matching for positive control spectra with a couple of other subsets of my NIST .msp library. Setting the match factor threshold very low, as Biswa advised, helps to produce good results. With the threshold at the default of 80% I didn't get matches for my positive control, even when that exact spectrum was in the library. When I lowered it to 5% I got matches, including the same spectrum scored at 100%. So there seems to be a bug with the way MS-FINDER returns database hits. I have attached the files to reproduce this.

Re: MS-DIAL and MS-FINDER library limits

Reply #4
Hi,

I am not sure if I understood your issue correctly. Today, I just downloaded your test query/library files, and analyzed them by ms-finder program. I attached the result and parameter setting used. Could you please let me know where the incorrect thing is?
Also, the source code is available at: http://prime.psc.riken.jp/compms/msfinder/main.html (click "Core"). Here you can check where the bug is.
Thanks,

Hiroshi

Re: MS-DIAL and MS-FINDER library limits

Reply #5
Tsugawa Hakase, kon'nichiwa.

Thank you for your response. I have spent quite some time working with MS-FINDER since you posted this response. I hoped to have worked out a solution myself but I am afraid I still do not understand how the software works to calculate match factors. I am not familiar with the C language so, while I had a quick look at the source code as you suggested, I could not easily identify the code where the spectral matching is carried out. 

I am still rather confused as to how MS-FINDER calculates spectral similarity. Is it a pure dot product score in the Structure Finder results window? Or is it weighted DP, or reverse DP? I am also confused as to how to get custom spectral libraries to produce consistent results in MS-FINDER. If I load the custom library I use for MS-DIAL I don't get any annotations from the Structure Finder. Instead I appended the contents of that file to the internal library file, MS-FINDER ver. 3.52 Windows\Resources\MSMS-DBs-vs1.etm The program now produces matches but these differ subtly from those assigned by MS-DIAL. Can you confirm whether the two programs use different versions of the dot product?