Skip to main content
Topic: How does MS-DIAL take representative spectra? (Read 315 times) previous topic - next topic

How does MS-DIAL take representative spectra?

Hello everyone,
Can anybody explain to me how does MS-DIAL take a representative spectrum for each aligned peaks? In my GC-MS project, I recently found that there are many peaks that remain unidentified possibly due to the representative spectra were not good. I say that because when I manually checked their spectra in AMDIS, many of them can be identified with a high match. I was thinking if the problem was the deconvolution process; however,  I have tried to use different sigma values (0.3, 0.5, and 0.7) for deconvolution and it did not improve too much. In addition, many of them have a unit peak in the chromatogram. As such, I prefer to know how does MS-DIAL takes the representative spectrum. Maybe I have too much samples for alignment, and MS-DIAL did not take the best spectrum for some peaks?

Thanks in advance.

best regards,
Qizhi Su

Re: How does MS-DIAL take representative spectra?

Reply #1
Hi Qizhi,

(below is also true in LC-MS/MS project.)
the representative spectrum is retrieved from one of the analysis files as follows.
1. the spectrum will be retrieved from a sample having the highest score of EI-MS spectral matching with a reference spectrum.
2. if there is no annotated information among samples, the spectrum will be retrieved from a sample having the highest peak height (by the defined quant mass in GC-MS project) among samples.

Can you find a good spectrum in each file?
If you find a good spectrum in each file but if you cannot find any good spectrum in the same/similar retention time area, there is a possibility that the alignment spot having such a good matched can be excluded in the finalization process as discussed in the below chat.
http://www.metabolomics-forum.com/index.php?topic=1391.msg4084#msg4084

(in such case, the issue may be resolved if you can define the quant mass field that does not share co-eluting metabolites each other.)

If you cannot find any good spectrum in any result file (in peak viewer), please send me a netcdf or abf file to check the issue.

Thanks,

Hiroshi


Re: How does MS-DIAL take representative spectra?

Reply #2
Yes, I can find good spectrum in the sample that has highest peak, but no good match in near retention time area. But sorry, I don't understand what do you mean "define the quant mass field that does not share co-eluting metabolites each other"? I am thinking if it is possible to keep only the alignment spot that has the highest match when there are many spots in similar retention time and same quan mass.

By the way, I would like to know how MS-DIAL calculates the "fill"? because I found that in many cases, the S/N ratio of a certain alignment spot in many samples is small that that in blanks, but not 0.  Will those samples be included in the "fill"? can we set that if a sample have S/N ratio lower than a certain value, they will not be regarded as positive samples, and their S/N or height can be set as null. In this way, we can know easily know which sample is truely positive/negative.

Re: How does MS-DIAL take representative spectra?

Reply #3
I recently found that untick "identification after alignment" can solve this problem. As I did identification before alignment, many of those unknown peaks were identified. 

Re: How does MS-DIAL take representative spectra?

Reply #4
It makes sense for me. Of course, using the option of "identification after alignment" is good when you have many samples and many reference records.... However, I recommend to do the identification process for all samples. I think, the result of peak alignment becomes better by this because the representative spectrum can be retrieved from a spot having a good match with the reference spectrum.

Hiroshi

Re: How does MS-DIAL take representative spectra?

Reply #5
yes, it is true. Many thanks.