Mass Spectral Libraries - Reproducibility in EI, API and Tandem Mass Spectrometry

Skip to Navigation

Education Article

  • Published: Apr 15, 2009
  • Author: Tony Mallet
  • Channels: Gas Chromatography / HPLC
thumbnail image: Mass Spectral Libraries - Reproducibility in EI, API and Tandem Mass Spectrometry

Professor Tony Mallet.
School of Science, University of Greenwich, Chatham Maritime, Kent ME4 4TB, UK.

A number of publications have recently appeared in which attempts have been made to determine the reproducibility of tandem mass spectra libraries. For several decades libraries of EI spectra have been used successfully for compound identification, especially from GC-MS data, while, in contrast, many difficulties seem to be present in trying to achieve the same success for tandem mass spectrometry data. This short essay attempts to answer the question of why this difference should occur, what is it about EI spectra which are so reproducible over a wide variety of instruments, laboratories and over long periods of time and what methods have been tried to improve the success rate of the latter.

EI Spectra

The time scales for electron ionisation ion formation, analysis, fragmentation and detection are illustrated in the diagram below:

(Adapted from E de Hoffman and V Stroobant, Mass Spectrometry 3rd edn 2007, p.274)

This scheme holds a key to the answers to the above questions and for a review of the early development of CID Cooks1 commentary is enlightening.

In an EI ion source the initial ionisation takes place in about 10-16 sec. The ion is electronically excited but no bonds can be broken or atoms rearranged at this time point. The excess electronic energy has to be reconfigured into vibrational and rotational kinetic energy and then has to be focussed into those bonds which are most likely to break and rearrange. No fragment ions can appear until roughly 10-10 sec has passed. At about 10-8 sec any excess electronic energy has been dissipated and intact molecular ions have returned to the ground state. The ions leave the ion source at about 10-6 sec. From now on they pass through the analyser and focussing devices and are finally detected but any further fragmentation or rearrangement will not be recorded on the ion detector no matter what analyser is used. Hence the nature of the EI mass spectrum is entirely defined inside the ion source and, as all EI sources are effectively equivalent, all EI spectra will have close similarities when recorded on instruments with varying analysers and dimensions.

Tandem Mass Spectra

Tandem mass spectra are produced by excitation of a preformed ion which has been extracted from the ion source. Methods include CID with an inert gas or surface, interaction with photons from a laser, or even thermal excitation. Two regimes of tandem mass spectrometry are commonly described. One is a high energy collision experiment where Kev ions are produced and the other is a low energy situation where the ions are formed with kinetic energies in the 10 to 100 ev range. High energy CID is performed in multi-sector magnetic analysers and in TOF-PSD and TOF/TOF instruments2, is usually the result of a single collision and is reported to start with an electronic excitation of the analyte ion. Low energy CID is more commonly reported and involves multiple collisions and direct excitation of bond vibration and rotation. For this reason alone it is likely that the mechanisms of ion fragmentation will differ considerably between these two regimes and this is well illustrated by the difference in product ion spectra of proteins, peptides and several other classes of compounds. The number of charges that the precursor ion holds will also influence the total kinetic energy it acquires on collision. One of the advantages of high over low energy CID is that amino acids undergo fragmentation at their side chains and this has been used to distinguish leucine from iso-leucine residues3.

Precursor to product ion transitions are unimolecular in nature and the two determining parameters which are the same for all such chemical reactions, i.e. the thermodynamics and the kinetics of the reaction. The CID low energy collision process takes place in about 10-10 sec but the total energy transferred to the precursor ion is hard to calculate on account of the multiple collisions taking place in a collision cell. In the case of tandem-in-time reactions even more collisions take place in ion traps than those of a tandem-in-time hexapole linear collision chamber. The transfer of energy is usually described in terms of the centre of mass collision energy defined as in the equation below:

Ecm = Elab x mg/(ma + mg)

Where Ecm and Elab are the centre of mass and laboratory energies, mg and ma are the masses of the collision gas and the analyte ion respectively. As the mass of the analyte ion increases the energy that can be transferred diminishes and some calculations have placed an upper limit of about 1200 dalton for effective CID. This does not take into account the multiple collision regime. An ion in a collision cell in a QqQ instrument may transit the cell in a few milliseconds and may undergo some 5 to 10 collisions. For a comprehensive review of the mechanisms of CID see the recent Mayer, Poon review4.

All of the above shows how a variety of experimental and design parameters are present which can alter the energies and reaction times of the fragmentation process.


EI spectral libraries are readily available from NIST and Wiley and include the 300,000 plus compound libraries as well as more specialist libraries focussing on drugs, steroids, fragrances, geochemicals and several others. Software for the refinement of GC/MS data is also available, some of it free, such as the AMDIS programme5, and they provide extra peak deconvolution and means for peak identification, which, combined with one of the above libraries, become a powerful tools for every day and automated compound identification of complex mixtures.

No such universal tandem MS library exists. The most successful implementations are restricted to in-house data sets which have been collated from one or two identical or very similar designs of mass spectrometer and sets of similar classes of analytes. Three groups have recently examined the variability of data from a spectrum of instrument designs and data processing implementations to seek for library identification from a set of MS/MS spectra.

A study was reported by Milman in 20056 in which a library of product ion spectra of 1743 compounds created from the entries in a number of databases from instruments with ion trap and triple quadrupole analysers. The collection included a number of replicate determinations of the same compound and several were taken at different collision energies. While searches made against this library from artificial 'unknowns' were able to achieve a fair degree of identification the results were unable to demonstrate that the two analyser types influenced the matching process.

A second set of results from Hopley et al.7 was obtained by 'round-robin' comparisons of MS/MS data produced following the initial; setting up of the instrument to produce virtually identical product ion spectra from a defined compound a so-called tuning point regime. The choice of tuning compound was reserpine and the tuning of the instrument was defined to obtain a fixed ratio (± 10%) of the abundance of a prominent product ion to that of the precursor ion. These tuning conditions were then applied during the analysis of 6 compounds in a blind fashion. These six were chosen from a collection of 48 compounds representing a variety of classes. Six ion traps, two quadrupole time of flight instruments, two triple quadrupoles and a hybrid triple quadrupole were used in the study. From the collection of 48 compounds a library of the product ion spectra of 44 compounds were produced.

The overall results showed that 31% of the spectra showed a 'no hit' on the library while 69% were correctly identified. Whereas in GC/MS EI library a fit of >90% is commonly obtained in similar tests the authors here used a more attainable 70% cut-off figure. The principal difficulties were observed when trap instruments were compared to 'in space' instruments where the pattern of fragmentation differed considerable from the latter configurations. Other problems included those which were a result of poor initial electrospray ionisation efficiency and varying detection levels of the precursor ion.

A third group who have been reporting on the problems in creating tandem mass spectrometry libraries is that of Oberacher. In a paper published in 20068 they described the creation of a library of the product ion spectra of 319 therapeutic drugs and drugs of illicit use. All spectra were recorded on a QqTOF instrument. Each compound was analysed using a range of collision energies and the accurate m/z values were also recorded. While many satisfactory matches were obtained for unknowns against this library the authors concluded that for acceptable forensic analysis it was also essential to include an effective gc separation prior to mass spectrometry. More recently9 this group has published two concurrent papers in which they have extended the above study. The library contained 3759 product ion spectra collected at a variety of collision energies and from a number of different instruments. These included QqTOF, QqLIT, QqQ, and LIT/FTICR analysers. The range of compounds chosen was again from a set of drug molecules together with a set of ergot alkaloids and the molecular weights ranged from about 150 to 450 dalton. They report the development of an efficient search algorithm regime which they claim permitted a 98% correct assignment rate.


Progress is being made in the creation and effectiveness of tandem mass spectral libraries and, while they do not yet approach the general usefulness of EI libraries, for defined sets of compound types and analysers they are being employed to simplify the task of interpreting the spectra of complex mixtures. In this task they will undoubtedly be enhanced by the increased separation performance of modern small particle HPLC columns and maybe in the future an approach similar to the AMDIS-GC/MS will appear for LC/MS analyses.


1. R G Cooks, J. Mass Spectrom. 1995, 30 1215-1221.
2. N V Gogichaeva, T Williams, M A Alterman, J. Am. Soc. Mass Spectrom. 2007, 18 279-284.
3. G Bouchoux et al., Org. Mass Spectrom. 1993, 28, 1064-1072.
4. P M Mayer, C Poon, Mass Spectrom. Reviews 2009, online pre-publication.
5. Available from:
6. B L Milman, Rapid Comm. Mass Spectrom. 2005, 19, 2833-2839.
7. C Hopley et al., Rapid Comm. Mass Spectrom. 2008, 22, 1779-1786.
8. M Pavlic, K Libiseller, H Oberacher, Anal. Biol. Chem. 2006, 386, 69-82.
9. H Oberacher et al., J. Mass Spectrom. 2009, 44, 485-493 and 494-502

Social Links

Share This Links

Bookmark and Share


Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Copyright Information

Interested in spectroscopy? Visit our sister site

Copyright © 2017 John Wiley & Sons, Inc. All Rights Reserved