Douglas B. Stauffer

Comparison of algorithms and databases for matching unknown mass spectra

The most used algorithms for the identification of electron-ionization mass spectra are INCOS and probability based matching (PBM). For unknown spectra of high purity, ∼75% of rank 1 answers are correct for both algorithms, matched against the National Institute of Standards and Technology 62,235 spectrum database. With matching criteria that retrieve 50% of the possible correct answers from the Wiley 228,998 spectrum database, 54% of the PBM and 42% of the INCOS answers are correct; for 85% purity unknowns, 48% and 27% are correct. For an unknown spectrum of two compounds, neither was reported in the first three INCOS answers; eight of the first ten PBM answers identify both components.

Comparative Evaluations of Mass Spectral Data Bases

Recent reports from the National Institute of Science and Technology (NIST) state that its large (53,994) collection of mass spectra is unique in “consisting almost entirely of complete spectra.” Our study of the 1989 Registry of Mass Spectral Data of 139,859 different spectra shows that its 53,994 spectra containing the most peaks average 108 peaks per spectrum, 48% larger than the NIST data base. Further, in matching unknown spectra of compounds present in both files, by using criteria yielding 68% reliability, 14% of the possible correct answers were recalled with the NIST data base versus 36% with the Registry.

An improved comprehensive data base for matching unknown mass spectra

Abstract The data base of reference mass spectra in computer-searchable form has been doubled in size to obtain 79560 spectra of 67128 different compounds. The utility of the new spectra was tested with the Probability Based Matching System using 392 unknowns selected at random from the original data base. The relative probability of incorrectly matching one of the new spectra is only half of that of the original spectra, demonstrating that the larger library is substantially less biased than the original. Of the best matching spectra, 74% were correct answers; when the data base was restricted to one spectrum of each compound, only 61% of the best matching spectra were correct.

An enlarged data base of electron-ionization mass spectra

The computer-searchable data base of reference mass spectra described earlier has been increased in size by 76%, so that it now contains 139,859 different spectra of 118,144 different compounds. The average number of peaks per spectrum is 53. All spectra were examined for errors by the Probability Based Matching (PBM) and the Quality Index (QI) algorithms and by human inspection. An improvement to the QI algorithm is based on the Terwilliger suggestion concerning saturated spectra. The number of different elemental compositions of compounds has increased by 64%. By using unknowns from the original data base with PBM, the probability that these incorrectly match a new spectrum is only 33% of that of incorrectly matching a spectrum in the original data base, further demonstrating that the variety of data in the library has been substantially expanded. Including additional reference spectra (measured under different conditions) of the same compound in the data base reduced the proportion of incorrect best-matching spectra by 42%.

Matching mass spectra against a large data base during GC/MS analysis

Abstract The data base of reference mass spectra has been expanded to contain 79,525 different spectra of 67,510 different compounds. The Probability Based Matching system for the identification of unknown mass spectra has been implemented on a commercial gas chromatograph/mass spectrometer (GC/MS) system for the identification of mass spectra of GC-separated components during the GC/MS run. Ordering of the reference file has made possible the search of a 41,000 spectra data base in 10 seconds during GC/MS analysis. For the interpretation of unknown mass spectra not in the reference file, the Self-Training Interpretive and Retrieval System has also been implemented on the same GC/MS instrument for off-line use.

Ion optics of multipole devices. I. Theory of the dodecapole

Abstract Properties are established for an ion-optical device consisting of twelve parallel rods placed equidistantly on a circle. In comparison with conventional devices, this dodecapole features improved deflection and quadrupole focusing of ion beams, two-directional focusing, correction of second- and third-order aberrations, and the simultaneous application of combinations of these properties.