Evaluation of mass spectral library search algorithms implemented in commercial software

An Integrated Workflow Assisted by In Silico Predictions To Expand the List of Priority Polycyclic Aromatic Compounds

The limited information in existing mass spectral libraries hinders an accurate understanding of the composition, behavior, and toxicity of organic pollutants. In this study, a total of 350 polycyclic aromatic compounds (PACs) in 9 categories were successfully identified in fine particulate matter by gas chromatography high resolution mass spectrometry. Using mass spectra and retention indexes predicted by in silico tools as complementary information, the scope of chemical identification was efficiently expanded by 27%. In addition, quantitative structure-activity relationship models provided toxicity data for over 70% of PACs, facilitating a comprehensive health risk assessment. On the basis of extensive identification, the cumulative noncarcinogenic risk of PACs warranted attention. Meanwhile, the carcinogenic risk of 53 individual analogues was noteworthy. These findings suggest that there is a pressing need for an updated list of priority PACs for routine monitoring and toxicological research since legacy polycyclic aromatic hydrocarbons (PAHs) contributed modestly to the overall abundance (18%) and carcinogenic risk (8%). A toxicological priority index approach was applied for relative chemical ranking considering the environmental occurrence, fate, toxicity, and analytical availability. A list of 39 priority analogues was compiled, which predominantly consisted of high-molecular-weight PAHs and alkyl derivatives. These priority PACs further enhanced source interpretation, and the highest carcinogenic risk was attributed to coal combustion.

Comparative Evaluation of Electron Ionization Mass Spectral Prediction Methods

During the past decade promising methods for computational prediction of electron ionization mass spectra have been developed. The most prominent ones are based on quantum chemistry (QCEIMS) and machine learning (CFM-EI, NEIMS). Here we provide a threefold comparison of these methods with respect to spectral prediction and compound identification. We found that there is no unambiguous way to determine the best of these three methods. Among other factors, we find that the choice of spectral distance functions play an important role regarding the performance for compound identification.

How searching against multiple libraries can lead to biased results in GC/MS-based metabolomics

RATIONALE

Databases of electron ionization mass spectra are often used in GC/MS-based untargeted metabolomics analysis. The results of the library search depend on several factors, such as the size and quality of the database, and the library search algorithm. We found out that the list of considered m/z values is another important parameter. Unfortunately, this information is not usually specified by software developers and it is hidden from the end user.

METHODS

We created synthetic data sets and figured out how several popular software products (AMDIS, ChromaTOF, MS Search, and Xcalibur) select the list of m/z values for the library search. Moreover, we considered data sets of real mass spectra (presented in both the NIST and FiehnLib libraries) and compared the library search results obtained within different software products. All programs under consideration use the NIST MS Search binaries to perform the library search using the Identity algorithm.

RESULTS

We found that AMDIS and ChromaTOF can give biased library search results under particular conditions. In untargeted metabolomics, this can happen when NIST and FiehnLib libraries are used simultaneously, the scan range of the instrument is less than 85, and the correct answer is present only in the FiehnLib library.

CONCLUSIONS

The main reason for biased results is that the information about the scan range is not stored in the metadata of library records. As a result, in the case of AMDIS and ChromaTOF software, some unrecorded peaks are considered as missing during the library search, the respective compound is penalized, and the correct answer falls outside the top five or even top 10 hits. At the same time, the default algorithm for selecting the list of considered m/z values implemented in MS Search is free from such unexpected behavior.

A general procedure for rounding m/z values in low-resolution mass spectra

It was revealed that nominal mass spectra extracted from the same NetCDF file using different gas chromatography/mass spectrometry (GC/MS) software products are not identical. This phenomenon is caused by differences in algorithms used for rounding floating-point m/z values to integers. It was found that all programs under consideration (AMDIS, ChemStation, ChromaTOF, MS Search, OpenChrom) use different procedures. It is necessary to know how fractional parts of accurate m/z values of ions are distributed to determine which algorithm yields more robust results. We estimated the respective distribution using two databases (PubChem and NIST). As a result, we came up with a procedure that minimizes the influence of random errors on rounding to integer m/z values. The procedure we suggest is to sum intensities of all floating-point m/z values in a bin [MZ - 0.38; MZ + 0.62] and assign MZ as a nominal m/z value, where MZ is an integer m/z value.

A procedure to detect and identify specific chemicals of potential inhalation toxicity concern in aerosols

OBJECTIVE

Understanding the potential inhalation toxicity of poorly characterized aerosols is challenging both because aerosols may contain numerous chemicals and because it is difficult to predict which chemicals may present significant inhalation toxicity concerns at the observed levels. We have developed a novel systematic procedure to address these challenges through non-targeted chemical analysis by two-dimensional gas chromatography-time-of-flight mass spectrometry (GC × GC-TOFMS) and assessment of the results using publicly available toxicity data to prioritize the tentatively identified detected chemicals according to potential inhalation toxicity.

MATERIALS AND METHODS

The procedure involves non-targeted chemical analysis of aerosol samples utilizing GC × GC-TOFMS, which is selected because it is an effective technique for detecting chemicals in complex samples and assigning tentative identities according to the mass spectra. For data evaluation, existing toxicity data (e.g. from the U.S. Environmental Protection Agency CompTox Chemicals Dashboard) are used to calculate multiple toxicity metrics that can be compared among the tentatively identified chemicals. These metrics include hazard quotient, incremental lifetime cancer risk, and metrics analogous to hazard quotient that we designated as exposure-(toxicology endpoint) ratios.

RESULTS AND DISCUSSION

We demonstrated the utility of our procedure by detecting, identifying, and prioritizing specific chemicals of potential inhalation toxicity concern in the mainstream smoke generated from the machine-smoking of marijuana blunts.

CONCLUSION

By designing a systematic approach for detecting and identifying numerous chemicals in complex aerosol samples and prioritizing the chemicals in relation to different inhalation toxicology endpoints, we have developed an effective approach to elucidate the potential inhalation toxicity of aerosols.

A quantitative reliability metric for querying large database

A redesigned quantitative reliability metric based on the F-distribution (QRMf) is reported for evaluating the reliability of library search. The QRMf provides orthogonal information to the comparison metric (e.g., dot product) and yields a probabilistic result. An intralibrary search can be considered as an idealized search because the top hit, i.e., the closest matching object, will match perfectly. If the search of an unknown object yields the same hit list as the intralibrary search, it would indicate good reliability. For each object in the hit list, a QRMf compares the order of an intralibrary and interlibrary search results and calculates a variance of interlibrary similarity metrics between the records of the intralibrary search and records in the corresponding positions of the interlibrary search. This variance that measures the discordance of the intra and interlibrary search can simply be compared to the variance of the similarity metrics within the interlibrary search results. The ratio of these variances follows an F-distribution that can be used to determine if the discordance is statistically significant and generates the probability based on the cumulative distribution function. The QRMf works for both similarity and dissimilarity and can be used for any queried object and comparison metric that is searched against a database. In this work, the QRMf was used along with the dot product similarity to query the mass spectra of novel synthetic opioids measured by gas chromatography-mass spectrometry (GC/MS). An automated pipeline was devised that used a basis set correction to assist peak detection. The basis was constructed by mass spectra obtained from the blank measurement preceding the analytical run to remove interferences from column bleed and septum degradation. After peak detection, the pipeline applied multivariate curve resolution to the chromatographic peak window to remove background components from the mass spectra. The corrected mass spectra were searched against a customized library for identification. The QRMf can be used along with the similarity metric to detect misidentifications and assist in finding the correct identification when it is not the closest match.

Spectral entropy outperforms MS/MS dot product similarity for small-molecule compound identification

Compound identification in small-molecule research, such as untargeted metabolomics or exposome research, relies on matching tandem mass spectrometry (MS/MS) spectra against experimental or in silico mass spectral libraries. Most software programs use dot product similarity scores. Here we introduce the concept of MS/MS spectral entropy to improve scoring results in MS/MS similarity searches via library matching. Entropy similarity outperformed 42 alternative similarity algorithms, including dot product similarity, when searching 434,287 spectra against the high-quality NIST20 library. Entropy similarity scores proved to be highly robust even when we added different levels of noise ions. When we applied entropy levels to 37,299 experimental spectra of natural products, false discovery rates of less than 10% were observed at entropy similarity score 0.75. Experimental human gut metabolome data were used to confirm that entropy similarity largely improved the accuracy of MS-based annotations in small-molecule research to false discovery rates below 10%, annotated new compounds and provided the basis to automatically flag poor-quality, noisy spectra.

Metabolomics-Driven Discovery of an Introduced Species and Two Malaysian Piper betle L. Variants

The differences in pungency of “sirih” imply the probable occurrence of several variants of Piper betle L. in Malaysia. However, the metabolite profiles underlying the pungency of the different variants remain a subject of further research. The differences in metabolite profiles of selected Malaysian P. betle variants were thus investigated; specifically, the leaf aqueous methanolic extracts and essential oils were analyzed via ¹H-NMR and GC-MS metabolomics, respectively. Principal component analysis (PCA) of the ¹H-NMR spectral data showed quantitative differences in the metabolite profiles of “sirih melayu” and “sirih india” and revealed an ambiguous group of samples with low acetic acid content, which was identified as Piper rubro-venosum hort. ex Rodigas based on DNA sequences of the internal transcribed spacer 2 (ITS2) region. The finding was supported by PCA of two GC-MS datasets of P. betle samples obtained from several states in Peninsular Malaysia, which displayed clustering of the samples into “sirih melayu” and “sirih india” groups. Higher abundance of chavicol acetate was consistently found to be characteristic of “sirih melayu”. The present research has provided preliminary evidence supporting the notion of occurrence of two P. betle variants in Malaysia based on chemical profiles, which may be related to the different genders of P. betle.

Simulation of the electron ionization mass spectra of the Novichok nerve agent

Novichok is one of the most feared and controversial nerve agents, which existence was confirmed only after the Salisbury attack in 2018. A new attack on August 2020, in Russia, was confirmed. After the 2018 attack, the agent was included in the list of the most dangerous chemicals of the Chemical Weapons Convention (CWC). However, information related to its electron ionization mass spectrometry (EI/MS), essential for unambiguous identification, is scarce. Therefore, investigations about Novichok EI/MS are urgent. In this work, we employed Born-Oppenheimer molecular dynamics through the Quantum Chemistry Electron Ionization Mass Spectrometry (QCEIMS) method to simulate and rationalize the EI/MS spectra and fragmentation pathways of 32 Novichok molecules recently incorporated into the CWC. The comparison of additional simulations with the measured EI spectrum of another Novichok analog is very favorable. A general scheme of the fragmentation pathways derived from simulation results was presented. The present results will be useful for elucidation and prediction of the EI spectra and fragmentation pathways of the dangerous Novichok nerve agent.

Chemical Identification and Confirmation of Contact Allergens

Identification of the etiological chemical agent(s) associated with a case(s) of allergic contact dermatitis (ACD) is important for both patient management and public health surveillance. Traditional patch testing can identify chemical allergens to which the patient is allergic. Confirmation of allergen presence in the causative ACD-associated material is presently dependent on labeling information, which may not list the allergenic chemical on the product label or safety data sheet. Dermatologists have expressed concern over the lack of laboratory support for chemical allergen identification and possibly quantification from patients' ACD-associated products. The aim of this review was to provide the clinician a primer to better understand the analytical chemistry of contact allergen confirmation and unknown identification, including types of analyses, required instrumentation, identification levels of confidence decision tree, limitations, and costs.

BMDMS-NP: A comprehensive ESI-MS/MS spectral library of natural compounds

The Bioinformatics & Molecular Design Research Center Mass Spectral Library - Natural Products (BMDMS-NP) is a library containing the mass spectra of natural compounds, especially plant specialized metabolites. At present, the library contains the electrospray ionization tandem mass spectrometry (ESI-MS/MS) spectra of 2739 plant metabolites that are commercially available. The contents of the library were made comprehensive by incorporating data generated under various experimental conditions for compounds with diverse molecular structures. The structural diversity of the BMDMS-NP data was evaluated using molecular fingerprints, and it was sufficiently exhaustive enough to represent the structures of the natural products commercially available. The MS/MS spectra of each metabolite were obtained with different types/brands of ion traps (tandem-in-time) or combinations of mass analyzers (tandem-in-space) at multiple collision energies. All spectra were measured repeatedly in each environment because variations can occur in spectra, even under the same conditions. Moreover, the probability, separability of searching, and transferability of this spectral library were evaluated against those of MS/MS libraries, namely: NIST17 and MoNA.

Deep Learning Driven GC-MS Library Search and Its Application for Metabolomics

Preliminary compound identification and peak annotation in gas chromatography-mass spectrometry is usually made using mass spectral databases. There are a few algorithms that enable performing a search of a spectrum in a large mass spectral library. In many cases, a library search procedure returns a wrong answer even if a correct compound is contained in a library. In this work, we present a deep learning driven approach to a library search in order to reduce the probability of such cases. Machine learning ranking (learning to rank) is a class of machine learning and deep learning algorithms that perform a comparison (ranking) of objects. This work introduces the usage of deep learning ranking for small molecules identification using low-resolution electron ionization mass spectrometry. Instead of simple similarity measures for two spectra, such as the dot product or the Euclidean distance between vectors that represent spectra, a deep convolutional neural network is used. The deep learning ranking model outperforms other approaches and enables reducing a fraction of wrong answers (at rank-1) by 9-23% depending on the used data set. Spectra from the Golm Metabolome Database, Human Metabolome Database, and FiehnLib were used for testing the model.

Molecular dynamics simulation of the electron ionization mass spectrum of tabun

Tabun (ethyl N,N-dimethylphosphoramidocyanidate), or GA, is a chemical warfare nerve agent produced during the World War II. The synthesis of its analogs is rather simple; thus, it is a significant threat. Furthermore, experiments with tabun and other nerve agents are greatly limited by the involved life risks and the severe restrictions imposed by the Chemical Weapons Convention. For these reasons, accurate theoretical assignment of fragmentation pathways can be especially important. In this work, we employ the Quantum Chemistry Electron Ionization Mass Spectra method, which combines molecular dynamics, quantum chemistry methods, and stochastic approaches, to accurately investigate the electron ionization/mass spectrometry (EI/MS) fragmentation spectrum and pathways of the tabun molecule. We found that different rearrangement reactions occur including a McLafferty involving the nitrile group. An essential and characteristic pathway for identification of tabun and analogs, a two-step fragmentation producing the m/z 70 ion, was confirmed. The present results will be also useful to predict EI/MS spectrum and fragmentation pathways of other members of the tabun family, namely, the O-alkyl/cycloalkyl N,N-dialkyl (methyl, ethyl, isopropyl, or propyl) phosphoramidocyanidates.

Predicting the absence of an unknown compound in a mass spectral database

Only a small subset of known organic compounds (amenable for gas chromatography/mass spectrometry) is present in the largest mass spectral databases (such as NIST or Wiley). Nevertheless, library search algorithms available in the market are not able to predict the absence of a compound in the database. In the present work, we have tried to implement such prediction by means of supervised classification. Training and validation set contained 1500 and 750 compounds, respectively. Two prediction sets (containing 750 and about 3000 mass spectra) were considered. The easiest-to-use models were built with only one input variable: match factor of the best candidate or InLib factor (both parameters were calculated within MS Search (NIST) software). Multivariate classification models were built by partial least squares discriminant analysis (PLS-DA); match factors of top n candidates were used as input variables. PLS-DA was found to be the most effective approach. The prediction efficiency strongly depended on the 'uniqueness' of mass spectra presented in the test set. PLS-DA model was able to correctly predict the absence of a compound in the database in 29.9% for prediction set #1 and in 74.4% for prediction set #2 (only 1.3% and 2.5% of compounds actually presented in the database were wrongly classified).

Steroidomics for highlighting novel serum biomarkers of testosterone doping

Aim: Quantification of testosterone (T) and 5α-dihydrotestosterone serum concentrations proved to be an efficient alternative to urinary steroid profiling for the detection of T doping. In this context, additional serum markers could be discovered by exploratory untargeted steroidomics studies. Results: Endogenous steroid metabolites were monitored by ultra high-performance liquid chromatography coupled to high-resolution mass spectrometry in serum samples collected during a T administration clinical trial. A three-step workflow for accurate review of annotation was used and multifactorial data analysis allowed highlighting promising serum biomarkers. Longitudinal monitoring of selected compounds was performed to assess T abuse detection capabilities. Conclusion: Application of serum steroidomics showed high potential for biomarker discovery of T doping, suggesting longitudinal monitoring of steroid hormones in serum as a significant improvement in detection of endogenous steroids abuse.

MetGem Software for the Generation of Molecular Networks Based on the t-SNE Algorithm

Molecular networking (MN) is becoming a standard bioinformatics tool in the metabolomic community. Its paradigm is based on the observation that compounds with a high degree of chemical similarity share comparable MS² fragmentation pathways. To afford a clear separation between MS² spectral clusters, only the most relevant similarity scores are selected using dedicated filtering steps requiring time-consuming parameter optimization. Depending on the filtering values selected, some scores are arbitrarily deleted and a part of the information is ignored. The problem of creating a reliable representation of MS² spectra data sets can be solved using algorithms developed for dimensionality reduction and pattern recognition purposes, such as t-distributed stochastic neighbor embedding (t-SNE). This multivariate embedding method pays particular attention to local details by using nonlinear outputs to represent the entire data space. To overcome the limitations inherent to the GNPS workflow and the networking architecture, we developed MetGem. Our software allows the parallel investigation of two complementary representations of the raw data set, one based on a classic GNPS-style MN and another based on the t-SNE algorithm. The t-SNE graph preserves the interactions between related groups of spectra, while the MN output allows an unambiguous separation of clusters. Additionally, almost all parameters can be tuned in real time, and new networks can be generated within a few seconds for small data sets. With the development of this unified interface ( https://metgem.github.io ), we fulfilled the need for a dedicated, user-friendly, local software for MS² comparison and spectral network generation.

Identification of small molecules using accurate mass MS/MS search

Tandem mass spectral library search (MS/MS) is the fastest way to correctly annotate MS/MS spectra from screening small molecules in fields such as environmental analysis, drug screening, lipid analysis, and metabolomics. The confidence in MS/MS-based annotation of chemical structures is impacted by instrumental settings and requirements, data acquisition modes including data-dependent and data-independent methods, library scoring algorithms, as well as post-curation steps. We critically discuss parameters that influence search results, such as mass accuracy, precursor ion isolation width, intensity thresholds, centroiding algorithms, and acquisition speed. A range of publicly and commercially available MS/MS databases such as NIST, MassBank, MoNA, LipidBlast, Wiley MSforID, and METLIN are surveyed. In addition, software tools including NIST MS Search, MS-DIAL, Mass Frontier, SmileMS, Mass++, and XCMS2 to perform fast MS/MS search are discussed. MS/MS scoring algorithms and challenges during compound annotation are reviewed. Advanced methods such as the in silico generation of tandem mass spectra using quantum chemistry and machine learning methods are covered. Community efforts for curation and sharing of tandem mass spectra that will allow for faster distribution of scientific discoveries are discussed.

Combining a Deconvolution and a Universal Library Search Algorithm for the Nontarget Analysis of Data-Independent Acquisition Mode Liquid Chromatography-High-Resolution Mass Spectrometry Results

Nontarget analysis is considered one of the most comprehensive tools for the identification of unknown compounds in a complex sample analyzed via liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS). Due to the complexity of the data generated via LC-HRMS, the data-dependent acquisition mode, which produces the MS² spectra of a limited number of the precursor ions, has been one of the most common approaches used during nontarget screening. However, data-independent acquisition mode produces highly complex spectra that require proper deconvolution and library search algorithms. We have developed a deconvolution algorithm and a universal library search algorithm (ULSA) for the analysis of complex spectra generated via data-independent acquisition. These algorithms were validated and tested using both semisynthetic and real environmental data. A total of 6000 randomly selected spectra from MassBank were introduced across the total ion chromatograms of 15 sludge extracts at three levels of background complexity for the validation of the algorithms via semisynthetic data. The deconvolution algorithm successfully extracted more than 60% of the added ions in the analytical signal for 95% of processed spectra (i.e., 3 complexity levels multiplied by 6000 spectra). The ULSA ranked the correct spectra among the top three for more than 95% of cases. We further tested the algorithms with 5 wastewater effluent extracts for 59 artificial unknown analytes (i.e., their presence or absence was confirmed via target analysis). These algorithms did not produce any cases of false identifications while correctly identifying ∼70% of the total inquiries. The implications, capabilities, and the limitations of both algorithms are further discussed.

Optimizing targeted/untargeted metabolomics by automating gas chromatography/mass spectrometry workflows

New database building and MS subtraction algorithms have been developed for automated, sequential two-dimensional gas chromatography/mass spectrometry (GC-GC/MS). This paper reports the first use of a database building tool, with full mass spectrum subtraction, that does not rely on high resolution MS data. The software was used to automatically inspect GC-GC/MS data of high elevation tea from Yunnan, China, to build a database of 350 target compounds. The database was then used with spectral deconvolution to identify 285 compounds by GC/MS of the same tea. Targeted analysis of low elevation tea by GC/MS resulted in the detection of 275 compounds. Non-targeted analysis, using MS subtraction, yielded an additional eight metabolites, unique to low elevation tea.

SANIST: optimization of a technology for compound identification based on the European Union directive with applications in forensic, pharmaceutical and food analyses

Electrospray Ionization and collision induced dissociation tandem mass spectrometry are usually employed to obtain compound identification through a mass spectra match. Different algorithms have been developed for this purpose (for example the nist match algorithm). These approaches compare the tandem mass spectra of the unknown analyte with the tandem mass spectra spectra of known compounds inserted in a database. The compounds are usually identified on the basis of spectral match value associated with a probability of recognition. However, this approach is not usually applied to multiple reaction monitoring transition spectra achieved by means of triple quadrupole apparatus, mainly due to the lack of a transition spectra database. The Surface Activated Chemical Ionization-Electrospray-NIST Bayesian model database search (SANIST) platform has been recently developed for new potential metabolite biomarker discovery, to confirm their identity and to use them for clinical and diagnostic applications. Here, we present an improved version of the SANIST platform that extends its application to forensic, pharmaceutical, and food analysis studies, where the compound identification rules are strict. The European Union (EU) has set directives for compound identification (EU directive 2002/657/EC). We have applied the SANIST method to identification of 11-nor-9-carboxytetrahydro-cannabinol in urine samples (an example of a forensic application), circulating levels of the immunosuppressive drug tacrolimus in blood (an example of a pharmaceutical application) and glyphosate in fruit juice (an example of a food analysis application) that meet the EU directive requirements. Copyright © 2016 John Wiley & Sons, Ltd.

How enhanced molecular ions in Cold EI improve compound identification by the NIST library

RATIONALE

Library-based compound identification with electron ionization (EI) mass spectrometry (MS) is a well-established identification method which provides the names and structures of sample compounds up to the isomer level. The library (such as NIST) search algorithm compares different EI mass spectra in the library's database with the measured EI mass spectrum, assigning each of them a similarity score called 'Match' and an overall identification probability. Cold EI, electron ionization of vibrationally cold molecules in supersonic molecular beams, provides mass spectra with all the standard EI fragment ions combined with enhanced Molecular Ions and high-mass fragments. As a result, Cold EI mass spectra differ from those provided by standard EI and tend to yield lower matching scores. However, in most cases, library identification actually improves with Cold EI, as library identification probabilities for the correct library mass spectra increase, despite the lower matching factors.

METHODS

This research examined the way that enhanced molecular ion abundances affect library identification probability and the way that Cold EI mass spectra, which include enhanced molecular ions and high-mass fragment ions, typically improve library identification results. It involved several computer simulations, which incrementally modified the relative abundances of the various ions and analyzed the resulting mass spectra.

RESULTS

The simulation results support previous measurements, showing that while enhanced molecular ion and high-mass fragment ions lower the matching factor of the correct library compound, the matching factors of the incorrect library candidates are lowered even more, resulting in a rise in the identification probability for the correct compound.

CONCLUSIONS

This behavior which was previously observed by analyzing Cold EI mass spectra can be explained by the fact that high-mass ions, and especially the molecular ion, characterize a compound more than low-mass ions and therefore carries more weight in library search identification algorithms. These ions are uniquely abundant in Cold EI, which therefore enables enhanced compound characterization along with improved NIST library based identification.