A New Concept for the Interpretation of Mass Spectra |
Based on a Combination of a Fragmentation Mechanism Database and a Computer Expert System Robert Mistrik HighChem, Ltd., Cajakova 18, 81105 Bratislava, Slovakia Overview To cope with the complexity of fragmentation processes, we have been developing a new concept for the interpretation of mass spectra. This makes use of a computer system that uses an intelligent fragmentation mechanism knowledge base for the prediction of unimolecular decomposition reactions. The knowledge base is encapsulated in a relational database that, in turn, is encapsulated in software with a graphical user interface. The software contains an expert system that automatically extracts the decomposition mechanism for each fragmentation reaction provided and determines the compound class range that the mechanism can be applied to. This approach is highly selective, which assures that skeletal and charge-remote rearrangements, ring closures and expansions, as well as compound specific mechanisms are applied to an appropriate structure. The purpose of this project is to apply database mechanisms to a user-provided structure and automatically propose fragmentation reactions for the given compound. Each proposed fragmentation step is connected to its template mechanism, which allows the source and original database entry used for prediction to be reviewed. The system supports the creation of user-defined fragmentation libraries that can be particularly useful in research involving generic structures exhibiting analogous fragmentation patterns. Introduction The interpretation of mass spectra requires an understanding of reactions in the gas phase. The biggest difficulty regarding the interpretation of mass spectra arises from the complexity of the chemistry of gaseous ions and the phenomena of unimolecular decompositions, which is not yet fully understood. The variability of reaction rates and ion stabilities is increased by the fact that ions can be generated by a wide variety of ionization methods, and mass-analyzed and detected by a wide range of instrumentation, which can be operated using a variety of experimental parameters. As a consequence, it is extremely difficult to propose wide-ranging fragmentation rules that are applicable to different kinds of chemical structures under different experimental conditions. Research into fragmentation patterns is, however, continuing and several more or less general fragmentation and rearrangement rules have been proposed1,2. Computer programs, such as Mass Frontier3, have been developed to facilitate spectra interpretation based on general fragmentation and rearrangement rules. HighChem Fragmentation Mechanism Library (FragLib™) We have been collating fragmentation mechanisms published in all the available printed media dedicated to mass spectrometry for a number of years (Table 1) and these have been entered into our computer system. Each reaction, along with the chemical structures, has been manually drawn in a full-featured reaction editor (FragEdit™) and saved in the relational database (FragBase™). Fragmentation pathways are accompanied with complementary information such as the ionization method, mass analyzer, ion activation and instrument, if this was available from the source (Figure 1).
To ensure the data is of the highest quality, fragmentation mechanisms have been rigorously evaluated in two stages: manual and automatic. The manual evaluation included accuracy and plausibility assessments of reaction mechanisms and consistency checking between fragment masses and peak m/z values, if the spectrum was available. The automatic evaluation included simple element, charge and radical consistency checks on both sides of the reaction, in addition to newly developed algorithms for complex electron mapping that revealed formally erroneous mechanisms. Numerous problems and errors regarding mechanisms were uncovered by both stages and either appropriate corrections were made, or these mechanisms were excluded from the library. The evaluation process was significantly slowed by the absence of unambiguous guidelines for the display of chemical structures and schemes showing fragmentation pathways in journal manuscript submission instructions. The authors often use unconventional and inconsistent indexing and abbreviations for substituents, isotopes, protonation sites, rearrangements, and neutral losses. As a result, approximately 3% of pathways were excluded due to the inability to semantically decode the depicted reactions or structures. The quality of published mechanisms is noticeably journal and author dependent and the overall error rate remains nearly constant over time. Table 1. FragLib™ Information Resources - Collection and Completion Time Line
Table 2. HighChem Fragmentation Mechanism Library (FragLib™) in Numbers
Automated Fragmentation Prediction Using FragLib™ At the core of the automated prediction of fragmentation and rearrangement pathways is a computer-based system for the extraction of unimolecular decomposition mechanisms from fragmentation schemes stored in FragLib. This complex software system decodes the underlying principle of fragmentation mechanisms from reaction drawings6 and builds a knowledge base of fragmentation events (Figure 3). This unique system works fully automatically and replaces the need for the tedious manual input of atom-atom correspondence in precursor and product ion pairs. A computer learning procedure was developed to allow the processing of specific fragmentation details. Similarly, in experimental mechanistic studies, labeled or generic structures can be used to direct the desired dissociation route. By means of deuteria or substituents (R) participation, the decoding algorithm unambiguously extracts the underlying mechanism (Figure 2). Since many reactions stored in FragLib follow general fragmentation rules which can be predicted by the use of different Mass Frontier’s modules, the library distinguishes between class specific mechanisms and general fragmentation reactions. After a reaction has been entered into the database, the system attempts to identify a general fragmentation rule and assigns the relevant reaction symbol above the arrow. In turn, if a compound specific mechanism is encountered, the reaction arrow is captioned with “Lib” (Fugure 4).
As regards fragmentation prediction, the determination of the preferred ionization site is as important as a detailed knowledge of the fragmentation mechanism. The system supports the virtual generation of charged molecules based on library ionization reactions using the exact location of the positive or negative charge, or the unspecified charge location symbol that can be assigned to a structure drawing. The extension of the charge localization concept, adopted in previous versions of Mass Frontier, to the processing of structures with an unspecified charge site represented a challenging task. The decoded database mechanisms can be applied to a user-provided structure for automated prediction of decomposition reactions. Any chemical structure drawn in structure editor can be fragmented utilizing the database knowledge. The computer system selects the appropriate fragmentation or rearrangement mechanisms, These are then consecutively employed in the generation of reactions and product ions (Figure 4).
The Software Fragmentation Library (FragLib™) is now an integral part of Mass Frontier software (Figure 4). Fragmentation pathways and algorithmically extracted mechanisms are stored in a relational database (FragBase™) that supports advanced name, author and substructure searches in addition to other functionalities. The database is managed via the graphical user interface, which contains a full-featured fragmentation reaction editor (FragEdit™). The system for the automated prediction of fragmentation and rearrangement pathways based on library mechanisms has been merged with a module that uses general fragmentation rules. The synergy of these two techniques, which make use of different principles, provides a powerful method of interpretation of mass spectra even when complicated rearrangements have had a significant influence on the overall dissociation process. This approach considerably increases the success rate when matching spectral peaks with generated fragments. The software allows users to build their own libraries with mechanisms that are related to compound classes frequently analyzed in their laboratories. This feature dramatically increases the efficiency of spectra interpretation since both general rules and the library mechanisms are able to explain the majority of spectral peaks in seconds.
Conclusion Fragmentation Library is a unique collection of several thousand mechanisms (Table 2). The mechanisms are stored in a knowledge base and can be automatically applied to any user-provided chemical structure to generate fragmentation pathways at an advanced level. Fragmentation Library together with predictive algorithms and data processing modules will be available in Mass Frontier 4.0 commercial software at the beginning of 2004. Work on enlarging the data collection is planned to continue until 2005, by which time we expect to have completed the screening of all the major sources of fragmentation mechanisms. References 1. McLafferty, F. W.; Turecek, F. Interpretation of Mass Spectra. USB Mill Valley, 1993
|