the confirmation of a match is best made by examination of the unknown spectrum and the library match. there are several things to keep in mind 1) the match score is the result of a mathematical calculation - like a matrix product. A high score shows more similarly, but to the mathematical method. 2) the libraries have only spectra for known compounds, and of those, only of those which have been recorded and included in a library. 3) there are many compounds which have spectra similar or identical to spectra of other compounds.

So, know your sample. this may help you to consider the possibility of a particular compound being present. when you consider a library hit suggested by the software, look to see if major ions present in the library hit are also present in the unknown (if a major ion is missing from the spectrum of the unknown, you don't have a match. And after you have given this thought and review, you have a best guess. Confirm the identity by comparison with authentic material or other information available to you.

Dear Khushman,

In general, when working with the NIST library, it is important to consider the subject of your search. The probability of a match will be dependent on the intensity of the signal, and the height and characteristics of background signals.

To identify a compound in an unknown sample with sufficient certainty, you need to verify not only the mass spectrum but also the retention time, so I would advise to check your spectra for likely compounds (based both on probability and on literature search on abundancy of compounds in similar samples, and on their retention behaviour), and then purchase the compounds you want to investigate and inject them as a standard to determine their retention time.

Kind regards,

Katrijn

the confirmation of a match is best made by examination of the unknown spectrum and the library match. there are several things to keep in mind 1) the match score is the result of a mathematical calculation - like a matrix product. A high score shows more similarly, but to the mathematical method. 2) the libraries have only spectra for known compounds, and of those, only of those which have been recorded and included in a library. 3) there are many compounds which have spectra similar or identical to spectra of other compounds.

So, know your sample. this may help you to consider the possibility of a particular compound being present. when you consider a library hit suggested by the software, look to see if major ions present in the library hit are also present in the unknown (if a major ion is missing from the spectrum of the unknown, you don't have a match. And after you have given this thought and review, you have a best guess. Confirm the identity by comparison with authentic material or other information available to you.

thank you for the reply Katrijn Van Huffel and Donald C Hilton. In my experiments i am not searching for a limited no. of specific compounds rather I am doing a sort of profiling experiment between a disease and its control. So to compare the two i need to find out what all compounds are coming in the two respective cases. This is the reason that i am not able to cross-check with the retention time and mass spectra collected from standards as i dont know which are the compounds that are significant to that particular disease. So that is the reason that i need to find the compound identity from the library matching and hence i was asking for the criteria to consider. Uptill now I am considering those compound to be taken into account which give me the match of more than 800 or 80% (not the probability%); of course after comparing the Mass Spectra between library hit and the unknown peak. Am i taking this correctly? or doing something wrong? If wrong then what should be my approach?

Hope you are getting my point.

Dear Khushman,

For comparing between samples, I would rather list the retention time, major mass fragments and peak area (and if available the name of the compound) for all detectable peaks (based on signal to noise ratio > 3). When selecting peaks for their match with the library, you would indeed miss out compounds not listed in the library, as Mr. Hilton suggested, or miss compounds where background signals are present, or identify the same compound differently in 2 samples.

When completing this list for disease and control samples, you will see similarities and differences between the lists. When a compound is present in one sample and missing in an other, I would filter the chromatogram on the main mass fragments to check if a small quantity might be present. Make sure to exclude compounds which are known artefacts of the sampling and separation materials you use.

Once you know what compounds can be relevant (I think you can find related literature to help you select relevant compounds), I would still consider purchasing pure compounds, not only to make your identification more grounded, but also to compare samples analysed on different moments, because the detector response can fluctuate, making quantitative comparison difficult. Use of an internal standard is also possible.

Best regards,

Katrijn

From my experience with ant extracts, I would say also that it helps much knowing more about your sample and the kinds of compounds in it. There is thus no set flawless approach for a specific sample. Mind that some compounds can elute at more or less the same time and thus their ions will superpose, often masquerading some results. In such case it is interesting to do runs with different protocols/columns, to note for any changes in the numbers and definitions of eluted peaks. Also knowing their chemical nature helps. For instance, when identifying hydrocarbons the library is useful to confirm it is a hydrocarbon and help interpret if it is linear, branched, or unsaturated -- its true ID will depend on Kovats index and comparison with standards. Also comparing with available literature and spectra will help much. Note that getting closest to being 100% sure about any ID will depend on the amount and diversity of the evidence pointing to that side, as everything in Science.

I would recomend you using an untargeted method to do the data analysis and then try to identify the relevant compounds (the ones that are different between disease and control) using all means necessary. See for example this paper and references therein for software suggestions: http://www.ncbi.nlm.nih.gov/pubmed/24656939

The best way to interpret GC-MS spectra is the match of their retention times with the previous reported ones if they are done in the same technical conditions..another thing in mass spectra is the m/z values...the presence of a particular m/z peaks shows the presence of that valued mass compound in it....

You can do it in this way:

1. see the retention times and then the mass spectra of that. if it matches then its same

2. if it doesn't then you directly see the mass spectra and then by seeing the parent peak and base peak you can think of the compounds reported or start with the basics of mass spectroscopy to understand the molecular structure of the compound. this link may help you in understanding the mass spectroscopy..

http://www.astbury.leeds.ac.uk/facil/MStut/mstutorial.htm

Good Luck,

Garima

I‘d like to introduce two articles to you. All the details about identification and quantification of GC-MS were contained in these issues. I suggest you to read them carefully and throughly. These articles really helped me a lot during my research. : )

1. "Compound Identification: A Journal of Agricultural and Food Chemistry Perspective"

2. “Quantitation of Sensory-Active and Bioactive Constituents of Food: A Journal of Agricultural and Food Chemistry Perspective”

published in 2007 and 2012 by ACS, respectively.

Factors, such as sampling, use of standards, identification methods, techniques of quantitation, and proper presentation and reporting of data, should be well considered at the initial stages of the investigation according to these standards. Hope everything goes well.

I'm having a similar problem with an untargeted analysis of fecal fatty acids in a novel animal (gorilla). Even more still, I am still unsure of how to treat my internal standard for quantitation, as well as the best way to approach deconvolution and integration.

Before doing library matching it is important that the data quality is as high as possible.

We are using PARAFAC2 for data processing of untargeted GC-MS data. We find that this approach provides superior deconvolution and clean MS-spectra with less noise which give better hits to NIST.

On our homepage www.msomics.com there is links to papers on PARAFAC2 as well as more information about our services regarding data processing of GC-MS data.

http://www.msomics.com

witout knowing the respons factor how could i compare to samples from GC/MS. if i do not know the concentrations in each sample ?. suppose we have 100comp for one sample and 70 for other but only 40 compounds common in between. The various area for one sample is independent between the compounds in that sample I assume?

best regards

Hi Khushman,

You should not identify compounds based purely on spectrum matches found when you search in mass-spectral libraries like NIST. Doing this will often lead to false identification (as you can find in many publications).

Don't forget you are using both a GC and a MS, so you should information from both to identify compounds. This means the retention information of the GC and the mass-spectrum of the MS. Combining these results will give a much more reliable identification.

There are no absolute limits for the match factor like below this match value it is an incorrect match and above that value it is a correct match. Ok, when the match factor is below 70 the you can safely say it's not a match. Above 70, everything is possible, false negatives below a match of 80 and false positives above a match of 90.

There is absolutely no limit value you can use for the probability. The probability value will tell you how unique you spectrum is within the NIST library. This is fixed value for a compound that can be anywhere between 5-100%. For toluene the probabilty will be somewhere between 55-65%, it will never be higher. For tetrachloroethene it will be between 90-100%. For compounds with isomers with similar spectra it will be around 100% divided by the number of isomers.

Whenever you make an identification keep track of the match factors, probabilities, SN, RI, boiling point, etc. When you have made 50, 100, 200 identifications learn from what these values mean for your samples.

Remember that it will take years to be an expert in identifying compounds using GCMS.

Regards, Alex