| Classifying Mass Spectra |
|
A simple example should help clarify how the classification of mass spectra can be carried out. Suppose we wanted to classify the derivatives of nicotine and caffeine. We have an EI spectrum of an unknown compound that we suspect is either a metabolite of nicotine or caffeine, and we need to find out which of these two compound classes it belongs to. To complete this task we need to retrieve sufficient amount of spectra for each group (e.g. nicotines and caffeines) which will then be submitted to a principal component analysis, to form clusters of each group. We can do this by performing a substructure search, where the search query will be the structure of nicotine and caffeine. If sufficient amount of spectra for each group are found (more than 10 spectra) you should select these records and add them, separately for each group, to the Spectra Classifier. Then you can assign an appropriate name to each group using the Selected Group of Spectra edit box. Feel free to use the copy (Ctrl C) and paste (Ctrl V) commands. Then, select a line in the left pane of Spectra Classifier and click on the Add button. Repeat this for the next group.Mass Frontier automatically assigns a graphic representation to each group. This is shown in the right pane for each line. You can easily change the graphic representations of groups using the Menu | Projection Layout menu item. Please note that the colors and type of graphic representation are only used to distinguish the groups in the plots and have absolutely no influence on the results of the analysis. Note: Spectral trees are supported by all the spectral modules except Spectra Classifier, which uses total composite spectra generated from trees.
When all the groups of spectra you want to classify are in the right window, click the Classify Now button to launch the classification process. After the generation is finished, a new Spectra Projector window will be displayed. We have retrieved two groups of spectra: nicotine and caffeine derivatives. We want to examine an unknown metabolite that we assume originated either from nicotine or caffeine. After we have prepared the two groups of spectra in Spectra Classifier, we can generate a Spectra Projector window by choosing the PCA option. The figure below shows that the two groups are separated into clusters. We can then add the unknown metabolite to the projection plane, by pasting or opening its spectrum into Spectra Projector. The projection of this external spectrum clearly shows that it belongs to the nicotine class (squares). This result can then be finally confirmed or rejected using the fragmentation pattern of nicotine.
|
