Matching of mixture mass spectra by subtraction of reference spectra

Abstract

technique is one of convenience since the fluorescence lifetime instrumentation can be used. This allows the cancellation of errors due to geometry or quenching in the calculation of the natural lifetime, T~, by T / @ Q = T~. The method does demonstrate that accurate quantum yields can be measured by using pulsed source excitation in the single photon counting mode.The method as presented here could benefit from certain instrumental improvements, such as monitoring the lamp intensity or a simultaneous measurement of absorbance and fluorescence. This would provide further precision and convenience by eliminating the manual change from absorbance to the fluorescence configuration. This could be accomplished in a manner similar to that of Holland et al. (5). This technique holds promise for the performance of routine quantum yield measurements and easy automation because of the intrinsic digital form of the data. A "spectrum-stripping" technique allows Improved identlflcation of minor mixture components in matching unknown mass spectra against a large reference file using the "Probabllity Based Matchlng" (PBM) system. The poor quantitative agreement expected between spectra of the same compound run under different condttions can be compensated by the hlgh information content of the mass values. Such spectrum subtraction Is complementary to the "reverse-search" feature of PBM, being especially valuable for identlfylng component spectra containing a number of peaks common to other components of the sample.

Probability-based-matching algorithm with forward searching capabilities for matching unknown mass spectra of mixtures

“…Final Algorithm Performance. Although qualitatively predicted by Atwater (10), the improvement to PBM performance for unknown spectra of mixtures is at least as impressive as that found for pure spectra. The 30% components showed (Figure 4) an average reduction of wrong answers of approximately 25%.…”

“…Probability based matching is probably the most widely used algorithm for retrieval from a comprehensive mass spectral data base (1,4,(8)(9)(10). This report extends preliminary correspondence (8) in describing the addition of forward searching capabilities to PBM applicable to unknown spectra of both pure compounds and mixtures utilizing automated subtraction of the best matching reference spectra from the unknown (10,11). An improvement to the "flagging" method (4) for removing spurious peaks also makes unnecessary the use of AK, the difference between the "confidence" (K) value found and that for a perfect match.…”

“…Application to Unknown Spectra of Mixtures. The spectrum subtraction technique of Atwater (10) offers a way to extract "pure" spectra of individual components from the unknown spectrum of a mixture. This assumes that each component spectrum, or a close match to it, is represented in the file; the recent doubling of the data base to contain mass spectra of 64376 different compounds (12,13) makes this probable for many types of unknowns.…”

“…To match the spectrum of an unknown whose purity is not known, the <25 best PBM matches retrieved with RLq >10% are saved (<35 and <45 retrieved PBM matches were also tested). The method of Atwater (10) was used to subtract the complete reference spectrum (in an amount based on the "% component value" and scaled) (9) of each retrieval from the unknown. All of the residual spectra so produced were then matched against each of the reference spectra originally retrieved (except the one subtracted), saving for each the largest RLp value found, designated RLa.…”

Estimating probabilities of correct identification from results of mass spectral library searches

“…The present statistics were obtained by using essentially complete spectra that are free of major impurities, so that reported probabilities strictly apply only to searches with spectra of comparable quality. Spectra having major impurity peaks would have led to less discriminating measures of probability, although this might be partly offset by using a more appropriate comparison logic ("reverse searching" for example [18]) or spectral subtraction methods [19].…”