Generating realistic data through modeling and parametric probability for the numerical evaluation of data processing algorithms in two-dimensional chromatography.

BACKGROUND: Comprehensive two-dimensional chromatography generates complex data sets, and numerous baseline correction and noise removal algorithms have been proposed in the past decade to address this challenge. However, evaluating their performance objectively is currently not possible due to a lack of objective data. RESULT: To tackle this issue, we introduce a versatile platform that models and reconstructs single-trace two-dimensional chromatography data, preserving peak parameters. This approach balances real experimental data with synthetic data for precise comparisons. We achieve this by employing a Skewed Lorentz-Normal model to represent each peak and creating probability distributions for relevant parameter sampling. The model's performance has been showcased through its application to two-dimensional gas chromatography data where it has created a data set with 458 peaks with an RMSE of 0.0048 or lower and minimal residuals compared to the original data. Additionally, the same process has been shown in liquid chromatography data. SIGNIFICANCE: Data analysis is an integral component of any analytical method. The development of new data processing strategies is of paramount importance to tackle the complex signals generated by state-of-the-art separation technology. Through the use of probability distributions, quantitative assessment of algorithm performance of new algorithms is now possible. Therefore, creating new opportunities for faster, more accurate, and simpler data analysis development.

Comprehensive Two-Dimensional Liquid Chromatography-High-Resolution Mass Spectrometry for Complex Protein Digest Analysis Using Parallel Gradients.

Despite the high gain in peak capacity, online comprehensive two-dimensional liquid chromatography coupled with high-resolution mass spectrometry (LC × LC-HRMS) has not yet been widely applied to the analysis of complex protein digests. One reason is the method's reduced sensitivity which can be linked to the high flow rates of the second separation dimension (2D). This results in higher dilution factors and the need for flow splitters to couple to ESI-MS. This study reports proof-of-principle results of the development of an RPLC × RPLC-HRMS method using parallel gradients (2D flow rate of 0.7 mL min-1) and its comparison to shifted gradient methods (2D of 1.4 mL min-1) for the analysis of complex digests using HRMS (QExactive-Plus MS). Shifted and parallel gradients resulted in high surface coverage (SC) and effective peak capacity (SC of 0.6226 and 0.7439 and effective peak capacity of 779 and 757 in 60 min). When applied to a cell line digest sample, parallel gradients allowed higher sensitivity (e.g., average MS intensity increased by a factor of 3), allowing for a higher number of identifications (e.g., about 2600 vs 3900 peptides). In addition, reducing the modulation time to 10 s significantly increased the number of MS/MS events that could be performed. When compared to a 1D-RPLC method, parallel RPLC × RPLC-HRMS methods offered a higher separation performance (FHWH from 0.12 to 0.018 min) with limited sensitivity losses resulting in an increase of analyte identifications (e.g., about 6000 vs 7000 peptides and 1500 vs 1990 proteins).

Studying Venom Toxin Variation Using Accurate Masses from Liquid Chromatography-Mass Spectrometry Coupled with Bioinformatic Tools.

This study provides a new methodology for the rapid analysis of numerous venom samples in an automated fashion. Here, we use LC-MS (Liquid Chromatography-Mass Spectrometry) for venom separation and toxin analysis at the accurate mass level combined with new in-house written bioinformatic scripts to obtain high-throughput results. This analytical methodology was validated using 31 venoms from all members of a monophyletic clade of Australian elapids: brown snakes (Pseudonaja spp.) and taipans (Oxyuranus spp.). In a previous study, we revealed extensive venom variation within this clade, but the data was manually processed and MS peaks were integrated into a time-consuming and labour-intensive approach. By comparing the manual approach to our new automated approach, we now present a faster and more efficient pipeline for analysing venom variation. Pooled venom separations with post-column toxin fractionations were performed for subsequent high-throughput venomics to obtain toxin IDs correlating to accurate masses for all fractionated toxins. This workflow adds another dimension to the field of venom analysis by providing opportunities to rapidly perform in-depth studies on venom variation. Our pipeline opens new possibilities for studying animal venoms as evolutionary model systems and investigating venom variation to aid in the development of better antivenoms.

Workflow to facilitate the detection of new psychoactive substances and drugs of abuse in influent urban wastewater.

The complexity around the dynamic markets for new psychoactive substances (NPS) forces researchers to develop and apply innovative analytical strategies to detect and identify them in influent urban wastewater. In this work a comprehensive suspect screening workflow following liquid chromatography - high resolution mass spectrometry analysis was established utilising the open-source InSpectra data processing platform and the HighResNPS library. In total, 278 urban influent wastewater samples from 47 sites in 16 countries were collected to investigate the presence of NPS and other drugs of abuse. A total of 50 compounds were detected in samples from at least one site. Most compounds found were prescription drugs such as gabapentin (detection frequency 79%), codeine (40%) and pregabalin (15%). However, cocaine was the most found illicit drug (83%), in all countries where samples were collected apart from the Republic of Korea and China. Eight NPS were also identified with this protocol: 3-methylmethcathinone 11%), eutylone (6%), etizolam (2%), 3-chloromethcathinone (4%), mitragynine (6%), phenibut (2%), 25I-NBOH (2%) and trimethoxyamphetamine (2%). The latter three have not previously been reported in municipal wastewater samples. The workflow employed allowed the prioritisation of features to be further investigated, reducing processing time and gaining in confidence in their identification.

Is Nontarget Analysis Ready for Regulatory Application? Influence of Peak-Picking Algorithms on Data Analysis.

The use of peak-picking algorithms is an essential step in all nontarget analysis (NTA) workflows. However, algorithm choice may influence reliability and reproducibility of results. Using a real-world data set, the aim of this study was to investigate how different peak-picking algorithms influence NTA results when exploring temporal and/or spatial trends. For this, drinking water catchment monitoring data, using passive samplers collected twice per year across Southeast Queensland, Australia (n = 18 sites) between 2014 and 2019, was investigated. Data were acquired using liquid chromatography coupled to high-resolution mass spectrometry. Peak picking was performed using five different programs/algorithms (SCIEX OS, MSDial, self-adjusting-feature-detection, two algorithms within MarkerView), keeping parameters identical whenever possible. The resulting feature lists revealed low overlap: 7.2% of features were picked by >3 algorithms, while 74% of features were only picked by a single algorithm. Trend evaluation of the data, using principal component analysis, showed significant variability between the approaches, with only one temporal and no spatial trend being identified by all algorithms. Manual evaluation of features of interest (p-value <0.01, log fold change >2) for one sampling site revealed high rates of incorrectly picked peaks (>70%) for three algorithms. Lower rates (<30%) were observed for the other algorithms, but with the caveat of not successfully picking all internal standards used as quality control. The choice is therefore currently between comprehensive and strict peak picking, either resulting in increased noise or missed peaks, respectively. Reproducibility of NTA results remains challenging when applied for regulatory frameworks.

Critical Assessment of the Chemical Space Covered by LC-HRMS Non-Targeted Analysis.

Non-targeted analysis (NTA) has emerged as a valuable approach for the comprehensive monitoring of chemicals of emerging concern (CECs) in the exposome. The NTA approach can theoretically identify compounds with diverse physicochemical properties and sources. Even though they are generic and have a wide scope, non-targeted analysis methods have been shown to have limitations in terms of their coverage of the chemical space, as the number of identified chemicals in each sample is very low (e.g., ≤5%). Investigating the chemical space that is covered by each NTA assay is crucial for understanding the limitations and challenges associated with the workflow, from the experimental methods to the data acquisition and data processing techniques. In this review, we examined recent NTA studies published between 2017 and 2023 that employed liquid chromatography-high-resolution mass spectrometry. The parameters used in each study were documented, and the reported chemicals at confidence levels 1 and 2 were retrieved. The chosen experimental setups and the quality of the reporting were critically evaluated and discussed. Our findings reveal that only around 2% of the estimated chemical space was covered by the NTA studies investigated for this review. Little to no trend was found between the experimental setup and the observed coverage due to the generic and wide scope of the NTA studies. The limited coverage of the chemical space by the reviewed NTA studies highlights the necessity for a more comprehensive approach in the experimental and data processing setups in order to enable the exploration of a broader range of chemical space, with the ultimate goal of protecting human and environmental health. Recommendations for further exploring a wider range of the chemical space are given.

Cumulative Neutral Loss Model for Fragment Deconvolution in Electrospray Ionization High-Resolution Mass Spectrometry Data.

Clean high-resolution mass spectra (HRMS) are essential to a successful structural elucidation of an unknown feature during nontarget analysis (NTA) workflows. This is a crucial step, particularly for the spectra generated during data-independent acquisition or during direct infusion experiments. The most commonly available tools only take advantage of the time domain for spectral cleanup. Here, we present an algorithm that combines the time domain and mass domain information to perform spectral deconvolution. The algorithm employs a probability-based cumulative neutral loss (CNL) model for fragment deconvolution. The optimized model, with a mass tolerance of 0.005 Da and a scoreCNL threshold of 0.00, was able to achieve a true positive rate (TPr) of 95.0%, a false discovery rate (FDr) of 20.6%, and a reduction rate of 35.4%. Additionally, the CNL model was extensively tested on real samples containing predominantly pesticides at different concentration levels and with matrix effects. Overall, the model was able to obtain a TPr above 88.8% with FD rates between 33 and 79% and reduction rates between 9 and 45%. Finally, the CNL model was compared with the retention time difference method and peak shape correlation analysis, showing that a combination of correlation analysis and the CNL model was the most effective for fragment deconvolution, obtaining a TPr of 84.7%, an FDr of 54.4%, and a reduction rate of 51.0%.

Novel Per- and Polyfluoroalkyl Substances Discovered in Cattle Exposed to AFFF-Impacted Groundwater.

The leaching of per- and polyfluoroalkyl substances (PFASs) from Australian firefighting training grounds has resulted in extensive contamination of groundwater and nearby farmlands. Humans, farm animals, and wildlife in these areas may have been exposed to complex mixtures of PFASs from aqueous film-forming foams (AFFFs). This study aimed to identify PFAS classes in pooled whole blood (n = 4) and serum (n = 4) from cattle exposed to AFFF-impacted groundwater and potentially discover new PFASs in blood. Thirty PFASs were identified at various levels of confidence (levels 1a-5a), including three novel compounds: (i) perfluorohexanesulfonamido 2-hydroxypropanoic acid (FHxSA-HOPrA), (ii) methyl((perfluorohexyl)sulfonyl)sulfuramidous acid, and (iii) methyl((perfluorooctyl)sulfonyl)sulfuramidous acid, belonging to two different classes. Biotransformation intermediate, perfluorohexanesulfonamido propanoic acid (FHxSA-PrA), hitherto unreported in biological samples, was detected in both whole blood and serum. Furthermore, perfluoroalkyl sulfonamides, including perfluoropropane sulfonamide (FPrSA), perfluorobutane sulfonamide (FBSA), and perfluorohexane sulfonamide (FHxSA) were predominantly detected in whole blood, suggesting that these accumulate in the cell fraction of blood. The suspect screening revealed several fluoroalkyl chain-substituted PFAS. The results suggest that targeting only the major PFASs in the plasma or serum of AFFF-exposed mammals likely underestimates the toxicological risks associated with exposure. Future studies of AFFF-exposed populations should include whole-blood analysis with high-resolution mass spectrometry to understand the true extent of PFAS exposure.

InSpectra - A platform for identifying emerging chemical threats.

Non-target analysis (NTA) employing high-resolution mass spectrometry (HRMS) coupled with liquid chromatography is increasingly being used to identify chemicals of biological relevance. HRMS datasets are large and complex making the identification of potentially relevant chemicals extremely challenging. As they are recorded in vendor-specific formats, interpreting them is often reliant on vendor-specific software that may not accommodate advancements in data processing. Here we present InSpectra, a vendor independent automated platform for the systematic detection of newly identified emerging chemical threats. InSpectra is web-based, open-source/access and modular providing highly flexible and extensible NTA and suspect screening workflows. As a cloud-based platform, InSpectra exploits parallel computing and big data archiving capabilities with a focus for sharing and community curation of HRMS data. InSpectra offers a reproducible and transparent approach for the identification, tracking and prioritisation of emerging chemical threats.

Instrument-independent chemometric models for rapid, calibration-free NPS isomer differentiation from mass spectral GC-MS data.

Chemometric analysis of mass spectral data for the purpose of differentiating positional isomers of novel psychoactive substances has seen a substantial increase in popularity in recent years. However, the process of generating a large and robust dataset for chemometric isomer identification is time consuming and impractical for forensic laboratories. To begin to address this problem, three sets of ortho/meta/para positional ring isomers (fluoroamphetamine (FA), fluoromethamphetamine (FMA), and methylmethcathinone (MMC)) were analyzed using multiple GC-MS instruments at three distinct laboratories. A diverse assortment of instrument manufacturers, model types, and parameters was utilized in order to incorporate substantial instrumental variation. The dataset was randomly split into 70% training and 30% validation sets, stratified by instrument. Following an approach based on Design of Experiments, the validation set was used to optimize the preprocessing steps performed prior to Linear Discriminant Analysis. Using the optimized model, a minimum m/z fragment threshold was determined to allow analysts to assess whether an unknown spectrum is of sufficient abundance and quality to be compared to the model. To assess the robustness of the models, a test set was developed utilizing two instruments from a fourth laboratory that was not involved in the generation of the primary dataset in addition to spectra from widely used mass spectral libraries. Of the spectra that reached the threshold, the classification accuracy was 100% for all three isomer types. Only two of the test and validation spectra that did not reach the threshold were misclassified. The results indicate that forensic illicit drug experts world-wide can use these models for robust NPS isomer identification on the basis of preprocessed mass spectral data without the need for acquiring reference drug standards and creating instrument specific GC-MS reference datasets. The continued robustness of the models could be ensured through international collaboration to collect data that captures all potential GC-MS instrumental variation encountered in forensic illicit drug analysis laboratories. This would allow every forensic institute to confidently assign isomeric structures without the need for additional chemical analysis.

Metabolome-Based Classification of Snake Venoms by Bioinformatic Tools.

Snakebite is considered a neglected tropical disease, and it is one of the most intricate ones. The variability found in snake venom is what makes it immensely complex to study. These variations are present both in the big and the small molecules found in snake venom. This study focused on examining the variability found in the venom's small molecules (i.e., mass range of 100-1000 Da) between two main families of venomous snakes-Elapidae and Viperidae-managing to create a model able to classify unknown samples by means of specific features, which can be extracted from their LC-MS data and output in a comprehensive list. The developed model also allowed further insight into the composition of snake venom by highlighting the most relevant metabolites of each group by clustering similarly composed venoms. The model was created by means of support vector machines and used 20 features, which were merged into 10 principal components. All samples from the first and second validation data subsets were correctly classified. Biological hypotheses relevant to the variation regarding the metabolites that were identified are also given.

Predicting RP-LC retention indices of structurally unknown chemicals from mass spectrometry data.

Non-target analysis combined with liquid chromatography high resolution mass spectrometry is considered one of the most comprehensive strategies for the detection and identification of known and unknown chemicals in complex samples. However, many compounds remain unidentified due to data complexity and limited number structures in chemical databases. In this work, we have developed and validated a novel machine learning algorithm to predict the retention index (r[Formula: see text]) values for structurally (un)known chemicals based on their measured fragmentation pattern. The developed model, for the first time, enabled the predication of r[Formula: see text] values without the need for the exact structure of the chemicals, with an [Formula: see text] of 0.91 and 0.77 and root mean squared error (RMSE) of 47 and 67 r[Formula: see text] units for the NORMAN ([Formula: see text]) and amide ([Formula: see text]) test sets, respectively. This fragment based model showed comparable accuracy in r[Formula: see text] prediction compared to conventional descriptor-based models that rely on known chemical structure, which obtained an [Formula: see text] of 0.85 with an RMSE of 67.

National Wastewater Reconnaissance of Analgesic Consumption in Australia.

A wastewater-based epidemiology (WBE) method is presented to estimate analgesic consumption and assess the burden of treated pain in Australian communities. Wastewater influent samples from 60 communities, representing ∼52% of Australia's population, were analyzed to quantify the concentration of analgesics used to treat pain and converted to estimates of the amount of drug consumed per day per 1000 inhabitants using pharmacokinetics and WBE data. Consumption was standardized to the defined daily dose per day per 1000 people. The population burden of pain treatment was classified as mild to moderate pain (for non-opioid analgesics) and strong to severe pain (for opioid analgesics). The mean per capita weighted total DDD of non-opioid analgesics was 0.029 DDD/day/person, and that of opioid-based analgesics was 0.037 DDD/day/person across Australia. A greater burden of pain (mild to moderate or strong to severe pain index) was observed at regional and remote sites. The correlation analysis of pain indices with different socioeconomic descriptors revealed that pain affects populations from high to low socioeconomic groups. Australians spent an estimated US $3.5 (AU $5) per day on analgesics. Our findings suggest that WBE could be an effective surveillance tool for estimating the consumption of analgesics at a population scale and assessing the total treated pain burden in communities.

From Molecular Descriptors to Intrinsic Fish Toxicity of Chemicals: An Alternative Approach to Chemical Prioritization.

The European and U.S. chemical agencies have listed approximately 800k chemicals about which knowledge of potential risks to human health and the environment is lacking. Filling these data gaps experimentally is impossible, so in silico approaches and prediction are essential. Many existing models are however limited by assumptions (e.g., linearity and continuity) and small training sets. In this study, we present a supervised direct classification model that connects molecular descriptors to toxicity. Categories can be driven by either data (using k-means clustering) or defined by regulation. This was tested via 907 experimentally defined 96 h LC50 values for acute fish toxicity. Our classification model explained ≈90% of the variance in our data for the training set and ≈80% for the test set. This strategy gave a 5-fold decrease in the frequency of incorrect categorization compared to a quantitative structure-activity relationship (QSAR) regression model. Our model was subsequently employed to predict the toxicity categories of ≈32k chemicals. A comparison between the model-based applicability domain (AD) and the training set AD was performed, suggesting that the training set-based AD is a more adequate way to avoid extrapolation when using such models. The better performance of our direct classification model compared to that of QSAR methods makes this approach a viable tool for assessing the hazards and risks of chemicals.

Chemometric Strategies for Fully Automated Interpretive Method Development in Liquid Chromatography.

The majority of liquid chromatography (LC) methods are still developed in a conventional manner, that is, by analysts who rely on their knowledge and experience to make method development decisions. In this work, a novel, open-source algorithm was developed for automated and interpretive method development of LC(-mass spectrometry) separations ("AutoLC"). A closed-loop workflow was constructed that interacted directly with the LC system and ran unsupervised in an automated fashion. To achieve this, several challenges related to peak tracking, retention modeling, the automated design of candidate gradient profiles, and the simulation of chromatograms were investigated. The algorithm was tested using two newly designed method development strategies. The first utilized retention modeling, whereas the second used a Bayesian-optimization machine learning approach. In both cases, the algorithm could arrive within 4-10 iterations (i.e., sets of method parameters) at an optimum of the objective function, which included resolution and analysis time as measures of performance. Retention modeling was found to be more efficient while depending on peak tracking, whereas Bayesian optimization was more flexible but limited in scalability. We have deliberately designed the algorithm to be modular to facilitate compatibility with previous and future work (e.g., previously published data handling algorithms).

Development of a comprehensive two-dimensional liquid chromatographic mass spectrometric method for the non-targeted identification of poly- and perfluoroalkyl substances in aqueous film-forming foams.

In this research, we developed an online comprehensive two-dimensional liquid chromatographic (LC × LC) method hyphenated with high-resolution mass spectrometry (HRMS) for the non-targeted identification of poly- and perfluorinated compounds (PFASs) in fire-fighting aqueous-film forming foams (AFFFs). The method exploited the combination of mixed-mode weak anion exchange-reversed phase with a octadecyl stationary phase, separating PFASs according to ionic classes and chain length. To develop and optimize the LC × LC method we used a reference training set of twenty-four anionic PFASs, representing the main classes of compounds occurring in AFFFs and covering a wide range of physicochemical properties. In particular, we investigated different modulation approaches to reduce injection band broadening and breakthrough in the second dimension separation. Active solvent and stationary phase assisted modulations were compared, with the best results obtained with the last approach. In the optimal conditions, the predicted peak capacity corrected for undersampling was higher than three-hundred in a separation space of about 60 min. Subsequently, the developed method was applied to the non-targeted analysis of two AFFF samples for the identification of homologous series of PFASs, in which it was possible to identify up to thirty-nine potential compounds of interest utilizing Kendrick mass defect analysis. Even within the samples, the features considered potential PFAS by mass defect analysis elute in the chromatographic regions discriminating for the ionic group and/or the chain length, thus confirming the applicability of the method presented for the analysis of AFFF mixtures and, to a further extent, of environmental matrices affected by the AFFF.

Collision Cross Section Prediction with Molecular Fingerprint Using Machine Learning.

High-resolution mass spectrometry is a promising technique in non-target screening (NTS) to monitor contaminants of emerging concern in complex samples. Current chemical identification strategies in NTS experiments typically depend on spectral libraries, chemical databases, and in silico fragmentation tools. However, small molecule identification remains challenging due to the lack of orthogonal sources of information (e.g., unique fragments). Collision cross section (CCS) values measured by ion mobility spectrometry (IMS) offer an additional identification dimension to increase the confidence level. Thanks to the advances in analytical instrumentation, an increasing application of IMS hybrid with high-resolution mass spectrometry (HRMS) in NTS has been reported in the recent decades. Several CCS prediction tools have been developed. However, limited CCS prediction methods were based on a large scale of chemical classes and cross-platform CCS measurements. We successfully developed two prediction models using a random forest machine learning algorithm. One of the approaches was based on chemicals' super classes; the other model was direct CCS prediction using molecular fingerprint. Over 13,324 CCS values from six different laboratories and PubChem using a variety of ion-mobility separation techniques were used for training and testing the models. The test accuracy for all the prediction models was over 0.85, and the median of relative residual was around 2.2%. The models can be applied to different IMS platforms to eliminate false positives in small molecule identification.

Socioeconomic status and public health in Australia: A wastewater-based study.

Analysis of untreated municipal wastewater is recognized as an innovative approach to assess population exposure to or consumption of various substances. Currently, there are no published wastewater-based studies investigating the relationships between catchment social, demographic, and economic characteristics with chemicals using advanced non-targeted techniques. In this study, fifteen wastewater samples covering 27% of the Australian population were collected during a population Census. The samples were analysed with a workflow employing liquid chromatography high-resolution mass spectrometry and chemometric tools for non-target analysis. Socioeconomic characteristics of catchment areas were generated using Geospatial Information Systems software. Potential correlations were explored between pseudo-mass loads of the identified compounds and socioeconomic and demographic descriptors of the wastewater catchments derived from Census data. Markers of public health (e.g., cardiac arrhythmia, cardiovascular disease, anxiety disorder and type 2 diabetes) were identified in the wastewater samples by the proposed workflow. They were positively correlated with descriptors of disadvantage in education, occupation, marital status and income, and negatively correlated with descriptors of advantage in education and occupation. In addition, markers of polypropylene glycol (PPG) and polyethylene glycol (PEG) related compounds were positively correlated with housing and occupation disadvantage. High positive correlations were found between separated and divorced people and specific drugs used to treat cardiac arrhythmia, cardiovascular disease, and depression. Our robust non-targeted methodology in combination with Census data can identify relationships between biomarkers of public health, human behaviour and lifestyle and socio-demographics of whole populations. Furthermore, it can identify specific areas and socioeconomic groups that may need more assistance than others for public health issues. This approach complements important public health information and enables large-scale national coverage with a relatively small number of samples.

Automated Feature Mining for Two-Dimensional Liquid Chromatography Applied to Polymers Enabled by Mass Remainder Analysis.

A fast algorithm for automated feature mining of synthetic (industrial) homopolymers or perfectly alternating copolymers was developed. Comprehensive two-dimensional liquid chromatography-mass spectrometry data (LC × LC-MS) was utilized, undergoing four distinct parts within the algorithm. Initially, the data is reduced by selecting regions of interest within the data. Then, all regions of interest are clustered on the time and mass-to-charge domain to obtain isotopic distributions. Afterward, single-value clusters and background signals are removed from the data structure. In the second part of the algorithm, the isotopic distributions are employed to define the charge state of the polymeric units and the charge-state reduced masses of the units are calculated. In the third part, the mass of the repeating unit (i.e., the monomer) is automatically selected by comparing all mass differences within the data structure. Using the mass of the repeating unit, mass remainder analysis can be performed on the data. This results in groups sharing the same end-group compositions. Lastly, combining information from the clustering step in the first part and the mass remainder analysis results in the creation of compositional series, which are mapped on the chromatogram. Series with similar chromatographic behavior are separated in the mass-remainder domain, whereas series with an overlapping mass remainder are separated in the chromatographic domain. These series were extracted within a calculation time of 3 min. The false positives were then assessed within a reasonable time. The algorithm is verified with LC × LC-MS data of an industrial hexahydrophthalic anhydride-derivatized propylene glycol-terephthalic acid copolyester. Afterward, a chemical structure proposal has been made for each compositional series found within the data.

Utilization of Machine Learning for the Differentiation of Positional NPS Isomers with Direct Analysis in Real Time Mass Spectrometry.

The differentiation of positional isomers is a well established analytical challenge for forensic laboratories. As more novel psychoactive substances (NPSs) are introduced to the illicit drug market, robust yet efficient methods of isomer identification are needed. Although current literature suggests that Direct Analysis in Real Time-Time-of-Flight mass spectrometry (DART-ToF) with in-source collision induced dissociation (is-CID) can be used to differentiate positional isomers, it is currently unclear whether this capability extends to positional isomers whose only structural difference is the precise location of a single substitution on an aromatic ring. The aim of this work was to determine whether chemometric analysis of DART-ToF data could offer forensic laboratories an alternative rapid and robust method of differentiating NPS positional ring isomers. To test the feasibility of this technique, three positional isomer sets (fluoroamphetamine, fluoromethamphetamine, and methylmethcathinone) were analyzed. Using a linear rail for consistent sample introduction, the three isomers of each type were analyzed 96 times over an eight-week timespan. The classification methods investigated included a univariate approach, the Welch t test at each included ion; a multivariate approach, linear discriminant analysis; and a machine learning approach, the Random Forest classifier. For each method, multiple validation techniques were used including restricting the classifier to data that was only generated on one day. Of these classification methods, the Random Forest algorithm was ultimately the most accurate and robust, consistently achieving out-of-bag error rates below 5%. At an inconclusive rate of approximately 5%, a success rate of 100% was obtained for isomer identification when applied to a randomly selected test set. The model was further tested with data acquired as a part of a different batch. The highest classification success rate was 93.9%, and error rates under 5% were consistently achieved.