On the risk of false positive identification using multiple ion monitoring in qualitative mass spectrometry: Large-scale intercomparisons with a comprehensive mass spectral library

Analysts involved in qualitative mass spectrometry have long debated the minimum data requirements for demonstrating that signals from an unknown sample are identical to those from a known compound. Often this process is carried out by comparing a few selected ions acquired by multiple ion monitoring (MIM), with due allowance for expected variability in response. In a few past experiments with electron-ionization mass spectrometry (EI-MS), the number of ions selected and the allowable variability in relative abundance were tested by comparing one spectrum against a library of mass spectra, where library spectra served to represent potential false positive signals in an analysis. We extended these experiments by carrying out large-scale intercomparisons between thousands of spectra and a library of one hundred thousand EI mass spectra. The results were analyzed to gain insights into the identification confidence associated with various numbers of selected ions. A new parameter was investigated for the first time, to take into account that a library spectrum with a different base peak than the search spectrum may still cause a false positive identification. The influence of peak correlation among the specific ions in all the library mass spectra was also studied. Our computations showed that (1) false positive identifications can result from similar compounds, or low-abundance peaks in unrelated compounds if the method calls for detection at very low levels; (2) a MIM method's identification confidence improves in a roughly continuous manner as more ions are monitored, about one order of magnitude for each additional ion selected; (3) full scan spectra still represent the best alternative, if instrument sensitivity is adequate. The use of large scale intercomparisons with a comprehensive library is the only way to provide direct evidence in support of these conclusions, which otherwise depend on the judgment and experience of individual analysts. There are implications for residue chemists who would rely on standardized confirmation criteria to assess the validity of a given confirmatory method. For example, standardized confirmation criteria should not be used in the absence of interference testing and rational selection of diagnostic ions.     

Computer identification of infrared spectra by correlation-based file searching

A new method for the computerized search and identification of infrared spectra has been developed and evaluated. Based on cross-correlation, the search system utilizes all spectral information in a digitized spectrum when it attempts to match an unknown spectrum to one in a small library of known spectra. To evaluate a spectral match, the search program calculates the cross-correlation function between the unknown and known (library) spectra which indicates their degree of similarity and allows library spectra to be ranked in order of probability of match to the unknown spectrum. In this study, several small infrared spectral libraries of structurally similar compounds were searched under conditions which examined the sensitivity of the search method to chemical and instrumental variations. Because the correlation technique is slower than conventional file-searching methods, it will probably find greatest use in the search of small collections of similar spectra or as a match-ranking procedure following preliminary selection by a faster search method.     

A correlation method in library search

The identification of unknown pure compounds or mixtures by means of mass spectral library search can be improved by partly resolving the spectra of the individual components within the spectrum of the measured unknown. This is accomplished by investigating sequential spectra in series of spectra; masses with highly correlating sequential intensities are clustered into individual groups. On the assumption that correlated masses belong to the same component spectrum, a filtering algorithm is developed to exclude spectra of non-identical compounds. The basic ideas, methods, examples and experiences gained with applications are reported.     

Towards a full reference library of MSn spectra. II: A perspective from the library of pesticide spectra extracted from the literature/Internet

To gain perspective on building full transferable libraries of MSⁿ spectra from their diverse/numerous collections, a new library was built from 1723 MS^>1 spectra (mainly MS² spectra) of 490 pesticides and related compounds. Spectra acquired on different types of tandem instruments in various experimental conditions were extracted from 168 literature articles and Internet sites. Testing of the library was based on searches where 'unknown' and reference spectra originated from different sources (mainly from different laboratories) were cross‐compared. The NIST 05 MS² library was added to the reference spectra. The library searches were performed with all the test spectra or were divided into different subsamples containing (a) various numbers of replicate spectra of test compounds or (b) spectra acquired from different instrument types. Thus, the dependence of true/false search (identification) result rates on different factors was explored. The percentage of 1^st rank correct identifications (true positives) for the only 'unknown' mass spectrum and two and more reference spectra and matching precursor ion m/z values was 89%. For qualified matches, above the cut‐off match factor, that rate decreased to 80%. The corresponding rates based on the best match for two and more 'unknown' and reference spectral replicates were 89–94%. For quadrupole instruments, the rates were even higher: 91–95% (one 'unknown' spectrum) and 90–100% (two and more such spectra). This study shows that MS² spectral libraries generated from the numerous literature/Internet sources are not less efficient for the goal of identification of unknown compounds including pesticides than very common EI‐MS¹ libraries and are almost as efficient as the most productive from current MS² spectral databases. Such libraries may be used as individual reference databases or supplements to large experimental spectral collections covering many groups of abundant compounds and different types of tandem mass spectrometers. Copyright © 2011 John Wiley & Sons, Ltd.     

Combined use of ESI–QqTOF-MS and ESI–QqTOF-MS/MS with mass-spectral library search for qualitative analysis of drugs

The potential of the combined use of ESI–QqTOF-MS and ESI–QqTOF-MS/MS with mass-spectral library search for the identification of therapeutic and illicit drugs has been evaluated. Reserpine was used for standardizing experimental conditions and for characterization of the performance of the applied mass spectrometric system. Experiments revealed that because of the mass accuracy, the stability of calibration, and the reproducibility of fragmentation, the QqTOF mass spectrometer is an appropriate platform for establishment of a tandem-mass-spectral library. Three-hundred and nineteen substances were used as reference samples to build the spectral library. For each reference compound, product-ion spectra were acquired at ten different collision-energy values between 5 eV and 50 eV. For identification of unknown compounds, a library search algorithm was developed. The closeness of matching between a measured product-ion spectrum and a spectrum stored in the library was characterized by a value called “match probability”, which took into account the number of matched fragment ions, the number of fragment ions observed in the two spectra, and the sum of the intensity differences calculated for matching fragments. A large value for the match probability indicated a close match between the measured and the reference spectrum. A unique feature of the library search algorithm—an implemented spectral purification option—enables characterization of multi-contributor fragment-ion spectra. With the aid of this software feature, substances comprising only 1.0% of the total amount of binary mixtures were unequivocally assigned, in addition to the isobaric main contributors. The spectral library was successfully applied to the characterization of 39 forensic casework samples. Electronic supplementary material Supplementary material is available in the online version of this article at and is accessible to authorized users.     

Beurteilungskriterien für Klassifikatoren zur automatischen Spektreninterpretation (pattern recognition)

In recent years numerous methods of pattern recognition have been tested for automatic interpretation of physicochemical data. Classifiers have been used successfully, especially with low resolution mass spectra. However, judgement of spectral classifiers (‘percentage of correctly classified spectra’) was often mathematically insufficiently defined. In this paper basic principles of the probability theory and information theory are used to derive objective criteria for binary classifiers. A classifier is an algorithm that uses a pattern vector (mass spectrum) and a priori probabilities for the classes (chemical structures) to which this vector belongs; the classification results are a posteriori probabilities for the classes. Predictive abilities for both classes or the information gain are suitable, objective criteria, to compare classifiers. Mathematical formulae are given and explained by examples from mass spectrometry.     

Evaluating electron ionization mass spectral library search results

As the size of various collections of electron ionization mass spectra gets larger, there is a continuing and increasing propensity to rely on the results of computerized library searches. The results of these computerized searches do not necessarily account for a spectrum that is produced by the mixture of two or more different compounds. Sometimes the submitted spectrum is not that of a compound whose spectrum is in the library. The quality of the spectrum submitted to the library is often such that the numerical confidence level reported for the search result is so low that it will cause the result to be disregarded. When the sample spectrum is matched against library spectra that have been condensed, the search result can be misleading. Three different examples of mass spectral search results are examined: one, with a high confidence level that the unknown has been identified, but the results are incorrect; one, where the spectrum of the unknown compound is not in the library; one, where consideration of the numerical search results would cause a positive identity not to be confirmed.     

Interpretation Automatique des Spectres de Masse de Composes Organiques. Mise au Point

Computerized methods for interpretation of mass spectra are increasing and many papers have now been published. Three different approaches can be used to identify an unknown compound. These methods are reviewed and discussed. First, the heuristic technique which tries to simulate the reasoning of the chemist and deduce the formulae. The only information needed are the mass spectra, the empirical formulae and the n.m.r. spectra. Some promising results have been obtained but the method has important restrictions. Another approach represents the mass spectra as points in a hyperspace. By developing decision surfaces it is possible to classify an unknown compound. This is called pattern recognition and the different empirical methods for recognizing patterns in mass spectral data are explained and reported. The last and simplest technique is to match an unknown spectrum against a library of standard references. When the unknown spectrum is in the library, this approach gives the best results.     

HPLC/APCI-FTICR-MS as a tool for identification of partial polar mutagenic compounds in effect-directed analysis

Identification of unknown compounds remains one of the biggest challenges for the assignment of adverse effects of sediment contamination and other complex environmental mixtures to responsible toxicants by effect-directed analysis (EDA). The identification depends on information gained from biotesting, chromatographic separation, and mass spectrometric detection. Thus, a methodology is provided for non-target identification of partial polar mutagenic polyaromatic compounds in sediment extracts by using polymeric reversed-phase HPLC column, high-resolution mass spectrometry and PubChem database. After visualization and processing the chromatogram constituents by using deconvolution software, the unambiguous elemental compositions generated were used as input in PubChem database to find a possible identity for the suspected species. The retrieved structures from the database search were refined by characterized chromatographic and mass spectrometric classifiers based on 55 model compounds comprising eight different classes representing mutagenic substructures. The applicability of the method was demonstrated by positive and tentative identification of constituents of mutagenic sediment fractions similar to selected model compounds.     

Libraries for spectrum identification: Method of normalized coordinates versus linear correlation

In this work it is proposed that an easy solution based directly on linear algebra in order to obtain the relation between a spectrum and a spectrum base. This solution is based on the algebraic determination of an unknown spectrum coordinates with respect to a spectral library base. The identification capacity comparison between this algebraic method and the linear correlation method has been shown using experimental spectra of polymers. Unlike the linear correlation (where the presence of impurities may decrease the discrimination capacity), this method allows to detect quantitatively the existence of a mixture of several substances in a sample and, consequently, to beer in mind impurities for improving the identification.     

Evaluation of the comparability of spectra generated using a tuning point protocol on twelve electrospray ionisation tandem-in-space mass spectrometers

Product ion spectra produced by collision-induced dissociation (CID) in tandem mass spectrometry can yield important structural information on organic compounds which can aid in their identification. However, differences in experimental conditions may have a strong effect on the degree of product ion formation and therefore on the features observed in product ion spectra. For this reason, a common approach for library building is the acquisition of several spectra, typically between 5 and 10, each at a different collision energy level. In this study, the use of an alternative approach was investigated, where a tuning point protocol was applied to tune the instruments in an attempt to standardise CID conditions prior to data acquisition. With this approach, the acquisition of a single mass spectrum was sufficient. The stability of the tuning point was investigated and the choice of a commercially available search package to assess spectral comparability was discussed. Finally, the product ion spectra of 33 compounds were acquired on twelve tandem-in-space instruments, including nine triple quadrupoles, one hybrid triple quadrupole linear ion trap and two quadrupole time-of-flight mass spectrometers, resulting in 2178 spectral comparisons being carried out. The results from the spectral comparisons suggest that the use of a tuning point enables the standardisation of the experimental conditions that affect the degree of product ion formation. Indeed, 84.5% of the comparisons demonstrated a good degree of spectral agreement with match scores greater than 700, which we believe is the minimum score for a tentative library match.     

MetFusion: integration of compound identification strategies

Mass spectrometry (MS) is an important analytical technique for the detection and identification of small compounds. The main bottleneck in the interpretation of metabolite profiling or screening experiments is the identification of unknown compounds from tandem mass spectra. Spectral libraries for tandem MS, such as MassBank or NIST, contain reference spectra for many compounds, but their limited chemical coverage reduces the chance for a correct and reliable identification of unknown spectra outside the database domain. On the other hand, compound databases like PubChem or ChemSpider have a much larger coverage of the chemical space, but they cannot be queried with spectral information directly. Recently, computational mass spectrometry methods and in silico fragmentation prediction allow users to search such databases of chemical structures. We present a new strategy called MetFusion to combine identification results from several resources, in particular, from the in silico fragmenter MetFrag with the spectral library MassBank to improve compound identification. We evaluate the performance on a set of 1062 spectra and achieve an improved ranking of the correct compound from rank 28 using MetFrag alone, to rank 7 with MetFusion, even if the correct compound and similar compounds are absent from the spectral library. On the basis of the evaluation, we extrapolate the performance of MetFusion to the KEGG compound database. Copyright © 2013 John Wiley & Sons, Ltd.     

Merits of random forests emerge in evaluation of chemometric classifiers by external validation

Real-world applications will inevitably entail divergence between samples on which chemometric classifiers are trained and the unknowns requiring classification. This has long been recognized, but there is a shortage of empirical studies on which classifiers perform best in ‘external validation’ (EV), where the unknown samples are subject to sources of variation relative to the population used to train the classifier. Survey of 286 classification studies in analytical chemistry found only 6.6% that stated elements of variance between training and test samples. Instead, most tested classifiers using hold-outs or resampling (usually cross-validation) from the same population used in training. The present study evaluated a wide range of classifiers on NMR and mass spectra of plant and food materials, from four projects with different data properties (e.g., different numbers and prevalence of classes) and classification objectives. Use of cross-validation was found to be optimistic relative to EV on samples of different provenance to the training set (e.g., different genotypes, different growth conditions, different seasons of crop harvest). For classifier evaluations across the diverse tasks, we used ranks-based non-parametric comparisons, and permutation-based significance tests. Although latent variable methods (e.g., PLSDA) were used in 64% of the surveyed papers, they were among the less successful classifiers in EV, and orthogonal signal correction was counterproductive. Instead, the best EV performances were obtained with machine learning schemes that coped with the high dimensionality (914–1898 features). Random forests confirmed their resilience to high dimensionality, as best overall performers on the full data, despite being used in only 4.5% of the surveyed papers. Most other machine learning classifiers were improved by a feature selection filter (ReliefF), but still did not out-perform random forests.     

Multidimensional computer evaluation of mass spectra

The increasing importance of spectroscopic methods as an analytical tool in industry, combined with the trend to automatize spectrometers, demands new standards in the quantity and quality of spectrum interpretation. Suitable computer programs should be able to predict structural features from mass spectral properties. The knowledge base is a structure-oriented mass spectral data collection consisting of some 42000 spectra and topologies. The comparison of selected mass spectral properties such as similarity, neutral losses and ion series of the unknown with the equivalent properties of the library spectra results in a set of corresponding structures. Subsequent substructure analysis yields a histogram of substructure frequencies containing information about their statistical relevance. The relevant substructure set may be recombined to produce a structure proposal, as is demonstrated for 1-acetyl-2-methoxy-4-trimethylsilyioxybenzene. In a second example, the relevant substructures derived by the interpretation system are used as input for the ¹³C-NMR substructure generator. This procedure reduces the solution space of the structure prediction algorithm considerably. Besides the spectrum interpretation, additional possibilities are available. The substructure search enables us, for example, to look for mass spectrometric reaction centres. Beyond that, substructure analysis is applicable to the determination of structural features typical of certain combinations of neutral losses and/or characteristic fragments.     

Information theory applied to feature selection of binary-coded infrared spectra for automated interpretation by retrieval of reference data

A method is described for feature selection from infrared spectra, intended for identification of organic compounds by computer-aided retrieval of reference data contained in small files. Complete discrimination of the binary-coded spectra is achieved by selecting a minimum number of spectral features; the information content is used as the selection criterion. The selection procedure is applied to five data sets (saturated and unsaturated hydrocarbons, alcohols, ethers and aldehydes/ketones) involving some 400 spectra. Each spectrum is uniquely coded by using about 10% of the 140 spectral features (binary-coded peak positions) available originally. For the intensity, a threshold of 50% appears to be applicable in some cases. For coding the frequency or wavelength parameter, wavenumbers (cm^-1) are preferred to wavelengths (mm). The method takes into account the a priori probabilities of spectral features and their correlations. Results of a retrieval program for a few “unknown” spectra are given.     

Discovery of false identification using similarity difference in GC–MS‐based metabolomics

Compound identification is a critical process in metabolomics. The widely used approach for compound identification in gas chromatography–mass spectrometry‐based metabolomics is spectrum matching, in which the mass spectral similarity between an experimental mass spectrum and each mass spectrum in a reference library is calculated. While various similarity measures have been developed to improve the overall accuracy of compound identification, little attention has been paid to reducing the false discovery rate. We, therefore, develop an approach for controlling the false identification rate using the distribution of the difference between the first and second highest spectral similarity scores. We further propose a model‐based approach to achieving a desired true positive rate. The developed method is applied to the National Institute of Standards and Technology mass spectral library, and its performance is compared with that of the conventional approach that uses only the maximum spectral similarity score. The results show that the developed method achieves a significantly higher F1 score and positive predictive value than did the conventional approach. Copyright © 2014 John Wiley & Sons, Ltd.     

Quality evaluation of tandem mass spectral libraries

Tandem mass spectral libraries are gaining more and more importance for the identification of unknowns in different fields of research, including metabolomics, forensics, toxicology, and environmental analysis. Particularly, the recent invention of reliable, robust, and transferable libraries has increased the general acceptance of these tools. Herein, we report on results obtained from thorough evaluation of the match reliabilities of two tandem mass spectral libraries: the MSforID library established by the Oberacher group in Innsbruck and the Weinmann library established by the Weinmann group in Freiburg. Three different experiments were performed: (1) Spectra of the libraries were searched against their corresponding library after excluding either this single compound-specific spectrum or all compound-specific spectra prior to searching; (2) the libraries were searched against each other using either library as reference set or sample set; (3) spectra acquired on different mass spectrometric instruments were matched to both libraries. Almost 13,000 tandem mass spectra were included in this study. The MSforID search algorithm was used for spectral matching. Statistical evaluation of the library search results revealed that principally both libraries enable the sensitive and specific identification of compounds. Due to higher mass accuracy of the QqTOF compared with the QTrap instrument, matches to the MSforID library were more reliable when comparing spectra with both libraries. Furthermore, only the MSforID library was shown to be efficiently transferable to different kinds of tandem mass spectrometers, including “tandem-in-time” instruments; this is due to the coverage of a large range of different collision energy settings—including the very low range—which is an outstanding characteristics of the MSforID library.     

An integrated method for spectrum extraction and compound identification from gas chromatography/mass spectrometry data

A method is presented for extracting individual component spectra from gas chromatography/mass spectrometry (GC/MS) data files and then using these spectra to identify target compounds by matching spectra in a reference library. It extends a published “model peak” approach which uses selected ion chromatograms as models for component shape. On the basis of this shape, individual mass spectral peak abundance profiles are extracted to produce a “purified” spectrum. In the present work, ion-counting noise is explicitly treated and a number of characteristic features of GC/MS data are taken into account. This allows spectrum extraction to be reliably performed down to very low signal levels and for overlapping components. A spectrum match factor for compound identification is developed that incorporates a number of new corrections, some of which employ information derived from chromatographic behavior. Test results suggest that the ability of this system to identify compounds is comparable to that of conventional analysis.     

Determination of ion structures in structurally related compounds using precursor ion fingerprinting

Structurally-related alkaloids were analyzed by electrospray ionization/multiple stage mass spectrometry (ESI/MS ⁿ ) at varying collision energies to demonstrate a conceptual algorithm, precursor ion fingerprinting (PIF). PIF is a new approach for interpreting and library-searching ESI mass spectra predicated on the precursor ions of structurally-related compounds and their matching product ion spectra. Multiple-stage mass spectra were compiled and constructed into “spectral trees” that illustrated the compounds’ product ion spectra in their respective mass spectral stages. The precursor ions of these alkaloids were characterized and their spectral trees incorporated into an MS ⁿ library. These data will be used to construct a universal, searchable, and transferable library of MS ⁿ spectra. In addition, PIF will generate a proposed structural arrangement utilizing previously characterized ion structures, which will assist in the identification of unknown compounds.     

Analysis of gamma-ray spectra from an aircraft-mounted array of high-resolution sensors

Gamma-ray spectra have been collected using the Environmental Radionuclide Sensor System.². Twenty aircraft-mounted, high-purity germanium sensors are utilized to collect high-resolution environmental spectra. Since time-over-target is limited for aerial surveys, one must often attempt to glean the maximum quantity of information from low-count spectra. The spectral data are collected in a time-stamped-list-mode, which time tags every gamma-ray. This allows a gamma-ray spectrum to be built without dilution for only the time-over-target. The analysis package utilizes photopeaks from natural background isotopes to autocalibrate and gain shift the individual spectra into a composite spectrum. The analysis software utilizes several unique techniques to robustly analyze low-count spectra. It carefully determines the spectral baseline, finds all the peaks which differ from the baseline by more than 4-standard deviations, uses a binary-search technique to fit Gaussian peaks, and utilizes a large library to identify peaks (including minor and escape peaks). Although in an aerial survey, the source geometry is often unknown or difficult to model, the software attempts to assign consistent source strengths to radionuclides. *** DIRECT SUPPORT *** A00E1009 00008