Retrieval The Probability Based Matching System

Satisfactory matching of an unknown and a reference spectrum constitutes the final proof of molecular identity (Section 5.5). However, evaluation of the reliability of such an identification must also consider the limitations of mass spectrometry peak-abundance agreement within a very small experimental error is often observed between spectra of stereo- and geometric isomers, and is possible between spectra of other closely similar compounds. The most reliable match is obtained by running the unknown and reference mass spectra under closely identical experimental conditions on the same instrument. However, the possible structures for the unknown must be reduced to a relatively small number of available reference compounds before their comparison is feasible experimentally. 

PBM incorporates the unique features of data weighting, proven invaluable for document retrieval from libraries (Salton 1975), and reverse search. The weighting involves the two principal types of data in low-resolution mass spectra masses and abundances. The probability that particular abundances will occur follows a log-normal distribution. The probability of occurrence of most mass values also varies in a predictable manner, as discussed under rules for 

Several features have been incorporated into the PBM system to minimize such inadequacies, and also those of sample and reference impurities. Speed improvements utilizing reference-file ordering now make it possible to search the 220,000 references in 3 s (Palisade 1991), fast enough to keep up with the chromatographic separation on most GC/MS runs. Thus it is recommended that PBM be run on each unknown before any examination by the interpreter. 

---