Identifying metabolites of new psychoactive substances using in silico prediction tools
NuMeCan Institute (Nutrition, Metabolisms and Cancer), CHU Rennes, Univ Rennes, INSERM, INRAE, UMR_A 1341, UMR_S 1317, 35000 Rennes, France
Clinical and Forensic Toxicology Laboratory, Rennes University Hospital, 35033 Rennes, France
Pharmacy Department, Rennes University Hospital, 35033 Rennes, France
Abstract
New psychoactive substances (NPS) pose an increasing challenge for clinical and forensic toxicology due to the initial lack of analytical and metabolic data. This study evaluates the performance of four in silico prediction tools (GLORYx, BioTransformer 3.0, SyGMa, and MetaTrans) in predicting the metabolism of seven NPS from five major chemical families (cathinones, synthetic cannabinoids, synthetic opioids, designer benzodiazepines, and dissociative anesthetics). The predicted metabolites were compared to those reported in the literature. The results revealed that SyGMa was the most exhaustive tool, predicting 437 metabolites, whereas MetaTrans predicted the fewest (61). GLORYx uniquely identified glutathione conjugation, while BioTransformer was particularly effective in predicting phase I reactions. However, no single tool provided complete predictions. Combining the four tools enabled the identification of several key biomarkers consistent with experimental data, such as m/z 238.1443 for eutylone and m/z 381.1926 for etonitazepipne. These findings highlight the need for integrated approaches to optimize metabolite prediction. Future advancements in artificial intelligence-based models could reduce false positives and enhance the accuracy of predictions, thus reinforcing the role of in silico tools in toxicological investigations.
Untitled section
Keywords: New psychoactive substances (NPS), In silico metabolism prediction, Biotransformation pathways, Toxicological biomarkers, Phase I and II metabolism, Prediction software comparison
Article notes
Untitled section
Received 2025 Jan 3; Accepted 2025 Mar 27; Issue date 2025.
Introduction
New psychoactive substances (NPS) comprise a diverse group of recreational drugs (including synthetic cannabinoids, arylcyclohexylamines, synthetic cathinones, new synthetic opiates, and designer benzodiazepines) engineered to replicate the pharmacological effects of illicit drugs such as cannabis, amphetamine, cocaine, 3,4-methylenedioxymethamphetamine, and lysergic acid diethylamide (UNODC 2023). These substances pose a significant challenge in toxicology due to the initial absence of analytical identification data regarding the parent molecule such as exact mass, MS/MS data and retention time. In addition, data on the metabolism of these substances is rarely available, although it is essential as it provides information on the metabolites that serve as consumption markers, particularly in cases where the parent molecule is no longer detected in biological matrices (Gicquel et al. 2024).
The biotransformation reactions of xenobiotics are divided into four main stages: entry of the xenobiotic (phase 0), functionalization reactions (phase I), conjugation reactions (phase II), and exit of the xenobiotic (phase III). In this article, we study the main phase I and II metabolites using NPS metabolism prediction software. NPS metabolism can be explored using a variety of approaches. Among these, human biological samples are considered the gold standard in metabolic studies because they offer real-life insights into the fate of parent compounds and their metabolites. Additionally, analyzing different types of samples provides information on the distribution of compounds and their metabolites in biological matrices. While blood and urine are the most commonly collected samples from living individuals, post-mortem samples can also be collected from bile, gastric contents, cardiac blood, or vitreous humor. The metabolites are known to accumulate particularly in urine and bile, though the limited availability of these samples poses a challenge to broader application (Bardal et al. 2011; Gicquel et al. 2024).
Metabolic studies in animals traditionally rely on rodent models to overcome limitations with human biological samples, but ethical considerations mandate the judicious use of animals (Pelletier et al. 2022a). Despite anatomical, physiological, and biochemical similarities to humans, significant interspecies differences in drug metabolism complicate the extrapolation of rodent, porcine, or canine data to humans (Lin 1995, 1998; Guengerich 1997; Bogaards et al. 2000; Dalgaard 2015).
In vitro studies have corroborated in vivo findings, identifying metabolites also present in human transformation reactions. Liver models, particularly genotyped primary human hepatocytes, are preferred for studying metabolism due to their expression of relevant enzymes and transporters, making them the gold standard despite limitations like high cost and variable enzyme expression (Gerets et al. 2012; Goncalves et al. 2022). Alternative models such as pooled human liver microsomes, pooled human S9 fractions, and differentiated HepaRG cells offer similar results in identifying major metabolites (Gicquel et al. 2024). Overall, several of these models have been employed in NPS metabolism studies using advanced analytical tools to reprocess data, particularly within non-targeted workflow and/or molecular networking (Allard et al. 2019; Pelletier et al. 2022b, 2023, 2024).
To go further, advancements in the understanding of metabolic mechanisms have facilitated the development of in silico metabolism prediction algorithms, which serve as convenient, open-access, time-efficient, and cost-effective tools for expanding metabolite searches and validating in vivo or in vitro data (Kirchmair et al. 2015). Various methodologies are employed in metabolism studies to create in silico systems, including (i) quantitative structure–activity relationship (QSAR) models, which posit that structurally similar molecules exhibit similar metabolic properties, (ii) quantum mechanical calculations for predicting reactivity, and (iii) docking simulations of potential substrates into enzyme active sites (Du et al. 2008; Gertrudes et al. 2012; Kazmi et al. 2019; Tyzack and Kirchmair 2019; Di Trana et al. 2021).
To date, very little data exist on the comparison of in silico xenobiotic metabolism prediction software in general (Boyce et al. 2023), and particularly on NPS. The aim of this study is to compare the results of various in silico metabolism prediction tools on seven NPS candidates belonging to the five most relevant chemical families to compare these tools in the metabolism studies of these new substances.
Materials and methods
In silico metabolite prediction
We included seven NPS belonging to the five chemical families most frequently reported in the literature and whose metabolism is accurately described in the literature (Salgueiro-Gonzalez et al. 2024; Santos et al. 2024): cathinones, dissociative anesthetics, synthetic cannabinoids, designer benzodiazepines, and new synthetic opiates (Table 1). The most relevant and up-to-date bibliographic reference was chosen for each NPS.
| Family | Name | Canonical SMILES code | Structures |
|---|---|---|---|
| Cathinones | Eutylone | CCC(C(= O)C1 = CC2 = C(C = C1)OCO2)NCC | |
| 4-Cl-PVP(4-chloro-pyrrolidinovalérophénone) | CCCC(C(= O)C1 = CC = C(C = C1)Cl)N2CCCC2 | ||
| Arylcyclohexylamine | 2F-DCK(2-fluoro-deschloro-kétamine) | CNC1(CCCCC1 = O)C2 = CC = CC = C2F | |
| New synthetic opioids (benzimidazoles derivative) | Etonitazepipne(N-Piperidinyl Etonitazene) | CCOC1 = CC = C(C = C1)CC2 = NC3 = C(N2CCN4CCCCC4)C = CC(= C3)[N+](= O)[O−] | |
| Designer Benzodiazepines | Adinazolam | CN(C)CC1 = NN = C2N1C3 = C(C = C(C = C3)Cl)C(= NC2)C4 = CC = CC = C4 | |
| (Hemi)Synthetic Cannabinoid | HHC(Hexahydrocannabinol) | CCCCCC1 = CC(= C2C3CC(CCC3C(OC2 = C1)(C)C)C)O | |
| ADB-Fubinaca | CC(C)(C)C(C(= O)N)NC(= O)C1 = NN(C2 = CC = CC = C21)CC3 = CC = C(C = C3)F |
The NPS metabolites were predicted using their SMILES string obtained in PubChem with four open-access software tools GLORYx, BioTransformer 3.0, SyGMa (Systematic Generation of potential Metabolites) and MetaTrans (Metabolite Translator). All programs predicting the biotransformation of xenobiotics uses reaction rules. They anticipate metabolic pathways and potential metabolites, including phase I and II reactions. However, they differ in their functionalities, which influences the results they deliver.
Among these tools, GLORYx is only available online (https://dev-nerdd.univie.ac.at/gloryx), and allow to export data on.sdf (Structure Data File). GLORYx combines site-of-metabolism prediction via learning models with reaction rules sets to predict and classify potential metabolite structures formed through phase I and/or phase II metabolism (de Bruyn et al. 2021). The “Phase 1 and phase 2 metabolism” parameter was used for GLORYx software.
BioTransformer 3.0 is available online (https://BioTransformer.ca/new) or as a downloadable, command-line only program. It is an open-access tool, enables rapid, accurate, and comprehensive prediction of small molecule metabolism in both mammals and environmental microorganisms (Djoumbou-Feunang et al. 2019). This software combines QSAR models with reaction rules to predict metabolites while identifying the enzymes involved. This feature adds an important biological dimension, facilitating integration with other biological data. For the analysis parameters, we used the ‘AllHuman’ mode, BioTransformer cycle number = 1 and Biotransformer CYP450 prediction Mode = 3.
MetaTrans and SyGMa software packages need to be installed on a computer. In this study, we predicted metabolites with the installed version of software packages using a bash script called Prediction_Metabo.sh (available on https://github.com/alexisbourdais/MetaTox/tree/origin). MetaTrans (Metabolites Translator) uses a deep learning architecture to anticipate metabolic reactions. Its approach relies on precise data, resulting in more targeted predictions. SyGMa (Systematic Generation of potential Metabolites) is a software package that predicts metabolites and biotransformation reactions using a reaction rule-based approach (Ridder and Wagener 2008). It uses predictive models and structural data to estimate the potential metabolites of a chemical compound. SyGMa stands out for its ability to predict a large number of metabolites, covering both phase I and phase II reactions. The parameters used for the SyGMa software are as follows: SyGMa Phase 1 cycle number = 1 and SyGMa Phase 2 cycle number = 1.
Comparison to literature data
A literature review was carried out on the PubMed, Google Scholar, and Web of Science databases to select studies presenting data on NPS metabolism. The following molecules were identified and selected for this study on the basis of their known metabolism in the literature: 2 synthetic cathinones (eutylone and 4-Cl-PVP), 1 dissociative anesthetic (2F-DCK), 1 synthetic cannabinoid (ADB-Fubinaca), 1 semi-synthetic cannabinoid (HHC), 1 designer benzodiazepine (adinazolam) and 1 new synthetic opiate (etonitazepipne). An overview of the methodology used in this study is shown in Fig. 1.
The lists of NPS metabolites presented in Tables 4, 5, 6, 7, 8, 9 and 10 were compiled by integrating metabolites from two sources: experimentally confirmed metabolites reported in the literature, and theoretical metabolites predicted by in silico tools. The tables below, therefore, present only those molecules found by the software and already described in the literature from relevant references that have studied the metabolism of the molecule of interest, using human samples and/or, where appropriate, in vitro models.
| Molecule Formula [M + H]+(m/z) | Mass shift | Chemical structure Biotransformation | GX | BT | MT | SM | In vivo (human blood) |
|---|---|---|---|---|---|---|---|
| EutyloneC13H17NO3236.1287 | – | – | – | – | – | – | |
| C12H17NO3224.1286 | − 12.0000 | X | X | X | X | X | |
| C13H19NO3238.1443 | + 2.0156 | X | X | X | X | X | |
| C13H19NO3238.1443 | + 2.0156 | X | X | X | X | X | |
| C13H15NO4250.1079 | + 13.9792 | X | X | ||||
| C12H17NO6S304.0854 | + 67.9567 | X | X | ||||
| C13H19NO6S318.1011 | + 81.9725 | X | X | ||||
| C18H25NO9400.1607 | + 164.032 | X | X | ||||
| C19H27NO9414.1764 | + 178.0477 | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Chemical structure Biotransformation | GX | BT | MT | SM | In vitro (HepaRG cells) | In vivo (human blood and urine) |
|---|---|---|---|---|---|---|---|---|
| 4-Cl-PVPC15H20ClNO 266.1312 | NA | – | – | – | – | – | – | |
| M4C15H18ClNO2280.1104 | + 13.9792 | X | X | X | X | |||
| M5C15H20ClNO2282.1260 | + 15.9948 | X | X | X | X | X | X | |
| M7C15H18ClNO3296.1053 | + 29.9741 | X | X | X | X | |||
| M8C15H20ClNO3298.1209 | + 31.9897 | X | X | X | ||||
| M5-GlucC21H28ClNO8458.1581 | + 192.0269 | X | X | |||||
| M7-GlucC21H26ClNO9472.1374 | + 206.0062 | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Structure Biotransformation | GX | BT | MT | SM | In vitro (HLMs/HepaRG cells) | In vivo (post-mortem samples) |
|---|---|---|---|---|---|---|---|---|
| 2F-DCKC13H16FNO222.1286 | NA | – | – | – | – | – | – | |
| Hydroxy-2F-DCKM01 or M03 or M19C13H16FNO2238.1238 | + 15.9952 | X | X | X | X | X/X | X | |
| Nor-2F-DCKM09C12H14FNO208.1130 | − 14.0156 | X | X | X | X | X/X | X | |
| Dihydro-2F-DCKM12 or M15C13H18FNO224.1442 | + 2.0156 | X | X | X | X/X | X | ||
| Hydroxy-nor-2F-DCK glucuronideM18C18H22FNO8400.1398 | + 178.0112 | X | X | |||||
| Hydroxy-2F-DCK glucuronideM20C19H24FNO8414.1558 | + 192.0272 | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Structure Biotransformation | GX | BT | MT | SM | In vitro (HLM) | In vivo (post-mortem urines) |
|---|---|---|---|---|---|---|---|---|
| EtonitazepipneC23H28N4O3409.2239 | – | – | – | – | – | – | – | |
| M1C23H30N4O379.2497 | − 29.9742 | X | X | X | X | X | ||
| M3C27H32N4O9557.2247 | + 148.0008 | X | X | |||||
| M4 or M8C21H24N4O4397.1876 | − 12.0363 | X | X | X | ||||
| M10C21H24N4O3381.1926 | − 28.0313 | X | X | X | X | X | X | |
| M11 or M21C23H28N4O4425.2188 | + 15.9949 | X | X | X | X | X | X | |
| M13 or M16 or M18C23H28N4O4425.2188 | + 15.9949 | X | X | X | X | X | X | |
| M14C29H36N4O10601.2509 | + 192.0270 | X | X | |||||
| M15C24H27N3O10518.1774 | + 108.9535 | X | X | |||||
| M17C23H28N4O5441.2138 | + 31.9899 | X | X | X | ||||
| M23C23H26N4O4423.2032 | + 13.9793 | X | X | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Structure Biotransformation | GX | BT | MT | SM | In vivo (human urines) |
|---|---|---|---|---|---|---|---|
| AdinazolamC19H18ClN5352.1329 | NA | - | - | - | - | - | |
| N-desmethyladinazolamC18H16ClN5338.1172 | − 14.0157 | X | X | X | X | X | |
| OH-alprazolamC17H13ClN4O325.0856 | − 27.0473 | X | X | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Structure Biotransformation | GX | BT | MT | SM | In vitro (human hepatocytes) | In vivo (human urines) |
|---|---|---|---|---|---|---|---|---|
| HHCC21H33O2317.2481 | – | – | – | – | – | – | – | |
| N1C27H40O8493.2801 | + 176.0320 | X | X | X | X | |||
| N2C27H40O9509.2750 | + 192.0269 | X | X | X | ||||
| N3C27H40O9509.2750 | + 192.0269 | X | X | X | ||||
| N6C27H38O10523.2543 | + 206.0062 | X | X | X | ||||
| N7C21H30O4347.2222 | + 29.9741 | X | X | X | X | X | ||
| N10C21H32O3333.2429 | + 15.9948 | X | X | X | X | X |
| Molecule Formula [M + H]+ (m/z) | Mass shift | Structure Biotransformation | GX | BT | MT | SM | In vitro (pooled human hepatocytes) |
|---|---|---|---|---|---|---|---|
| ADB-FubinacaParentC21H23FN4O2383.1883 | NA | - | - | - | - | - | |
| M14C21H23N4O3F399.1832 | + 15.9949 | X | X | X | X | X | |
| M16C21H23N4O3F399.1832 | + 15.9949 | X | X | X | X | X | |
| M3C14H18N4O2275.1509 | − 108.0374 | X | X | X | X | ||
| M11C21H23N4O4F415.1791 | + 31.9898 | X | X | ||||
| M10C27H31N4O9F575.2156 | + 192.0263 | X | X | ||||
| M18C27H30N3O9F560.2044 | + 177.0161 | X | X | X | |||
| M22C21H22N3O3F384.1726 | + 0.9843 | X | X | X |
Chemical structures drawing
All the molecules were drawn using Chemdraw 23.1.1 software. To standardize the structures, we used the ACS mode of document 1996.
Results
Discussion
In this study, we aimed to compare the results of four in silico metabolism prediction software packages on 7 NPS candidates belonging to five different chemical families to better position these tools in the metabolism studies of these new substances. Indeed, in silico prediction software are useful tools to explore the metabolism of NPS (Pelletier et al. 2022a, 2024). By combining biotransformation reaction modeling with databases and learning algorithms, these software packages offer innovative perspectives for describing potential metabolites, thus reducing the time and costs associated with in vitro and in vivo studies. Also these software packages are freely available and easy to use, despite their performance varies considerably depending on the software used.
GLORYx predicted a total of 191 metabolites, covering both phases I and II biotransformations. Interestingly, GLORYx proposed less common predictions, such as glutathione conjugation for Adinazolam and 4-Cl-PVP, demonstrating its potential for exploring complex metabolic pathways. Lastly, while GLORYx proposes ranks and scores to prioritize metabolic scenarios and support metabolite identification performed in vitro or in vivo as already used in the literature (Di Trana et al. 2021; Berardinelli et al. 2022, 2024b; Pelletier et al. 2022b; Brunetti et al. 2023), we included in this study all the proposed metabolites to be able to compare the software packages with each other.
With 91 predicted metabolites, BioTransformer 3.0 positions itself as a balanced tool, predicting major reactions such as hydroxylation and dealkylation for most of the evaluated NPS. In particular, this advantage has been used in the literature to add the expected main metabolites to the inclusion lists during analysis by high-performance liquid chromatography–high-resolution mass spectrometry (Verougstraete et al. 2023). However, BioTransformer sometimes lacks precision in certain predictions. For instance, it failed to identify glucuroconjugated metabolites for eutylone and 2F-DCK, even though these are often essential biomarkers in toxicological studies (Gicquel et al. 2021; Pelletier et al. 2023).
MetaTrans predicted the fewest number of metabolites (n = 80), without including any phase II reactions for the evaluated substances. This limitation reduces its relevance for comprehensive exploration of NPS metabolism. Nevertheless, MetaTrans proved its value by predicting certain key metabolites, such as Nor-2F-DCK for 2F-DCK, a relevant biotransformation identified in the literature (Gicquel et al. 2021). This demonstrates that even with a limited number of results, the quality of the predictions remains significant.
With 437 predicted metabolites for the seven NPS studied, SyGMa offers the broadest range, particularly for conjugation reactions such as glucuronidation and sulfation. However, this advantage can also be a drawback when the software predicts redundant or aberrant metabolites. In the 2F-DCK example, SyGMa proposed 10 metabolites but only three unique biotransformations, complicating data analysis. Its exhaustiveness can thus result in additional work to filter relevant results.
In terms of overall coverage, SyGMa emerges as the most exhaustive tool, although this can result in superfluous predictions. MetaTrans and BioTransformer focus more on precise but limited predictions, which can be advantageous for targeted studies. GLORYx, with its balance between coverage and relevance, offers an interesting solution based on these results.
Our findings indicate that SyGMa is particularly effective in generating phase II metabolites for all tested molecules. However, combining multiple tools significantly enhanced the metabolite coverage. To predict reliable biomarkers in silico, the most robust approach involves integrating several software tools, as demonstrated in previous studies (Pelletier et al. 2022a, b, 2024). This integrative approach minimizes aberrant metabolites and streamlines the selection of relevant candidates from extensive predictions. By addressing individual inconsistencies, the combined use of the four software tools identifies metabolites with the highest likelihood of occurring in vivo when consistently predicted across all tools (Gicquel et al. 2024).
As part of this study’s limitations, it is likely that certain metabolites identified in silico are indeed present in biological samples but were not detected in the studies. It would, therefore, be appropriate to carry out other in vitro or in vivo analyses using the in silico data generated here. Conversely, we show here that in silico software cannot always suggest metabolites of interest. In the example of adinazolam, the software predicts between 9 and 69 metabolites. In the literature (Fraser et al. 1993), four metabolites were identified as being of biomarker interest, but only two of these were suggested by the software, leaving 50% of the relevant metabolites unidentified. N,N-didesmethyladinazolam is not proposed, although it seems relevant that it should be found, given the chemical structure of this benzodiazepine. Among other drug classes, N,N-didesmethyl-derivatives are notably found with tramadol and venlafaxine.
With regard to synthetic cathinones (eutylone and 4-Cl-PVP), some metabolites identified as potential consumption markers in the literature were predicted by the four software packages, such as m/z 238.1443 for eutylone and M5 for 4-Cl-PVP. These results might suggest that the prediction of metabolites of other cathinones by these four software packages could lead to the proposal of relevant consumption markers to look for even before obtaining in vivo or in vitro samples. Similar results were obtained with three other molecules for which the metabolites predicted in common by the four software packages constitute proposed n markers, such as M09 for 2F-DCK, M10 for etonitazepipne and M14/M16 for ADB-Fubinaca, in vitro or in vivo consumption.
Overall, although this work enables us to evaluate the performance of four metabolic prediction software packages, we show that these tools cannot yet replace in vitro and in vivo experiments to identify relevant biomarkers. Looking ahead, the development of next-generation metabolic prediction tools with lower false-positive rates could further refine predictions, enhancing the identification of biomarkers in the absence of in vitro or in vivo studies. Promising examples include Metapredictor (Zhu et al. 2024) and Semeta (https://optibrium.com/products/semeta/) whose metabolite prediction should be studied in the light of the literature. Future advancements in machine learning and the integration of artificial intelligence are thus expected to yield more accurate and specific predictions.
Conclusion
In silico prediction tools offer a valuable, cost-effective, and efficient approach to explore the metabolism of new psychoactive substances (NPS), particularly in the absence of biological samples. SyGMa demonstrated the most comprehensive coverage, especially in predicting phase II metabolites, while BioTransformer 3.0 and MetaTrans provided more targeted but narrower predictions. GLORYx stood out for its innovative pathways, such as glutathione conjugation, although these conjugates have not been identified in the literature. Combining multiple tools proved critical to overcoming individual limitations and enhancing prediction reliability, as demonstrated by the consistent identification of key metabolites such as m/z 238.1443 for eutylone and m/z 381.1926 for etonitazepipne. This integrative approach increases the likelihood of identifying reliable biomarkers of consumption. Looking forward, the integration of artificial intelligence and machine learning into next-generation tools promises to improve accuracy, reduce false positives, and better predict complex metabolic pathways. By bridging the gaps in current methods, in silico tools are poised to play a pivotal role in toxicological investigations, supporting early detection and characterization of emerging NPS.
Funding
Open access funding provided by Centre Hospitalier Universitaire de Rennes.
Declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Footnotes
Footnote Group
References
Untitled section
References
- Allard S, Allard P, Morel I, Gicquel T (2019) Application of a molecular networking approach for clinical and forensic toxicology exemplified in three cases involving 3-MeO-PCP, doxylamine, and chlormequat. Drug Test Anal 11:669–677. 10.1002/dta.2550
- Bardal SK, Waechter JE, Martin DS (2011) Pharmacokinetics. In: Applied pharmacology. Elsevier, pp 17–34
- Berardinelli D, Montanari E, Malaca S et al (2022) 4-Hydroxy-N, N-methylpropyltryptamine (4-OH-MPT) in vitro human metabolism. Toxicol Analyt Clin 34:S96. 10.1016/j.toxac.2022.06.147
- Berardinelli D, Taoussi O, Carlier J et al (2024a) In vitro, in vivo metabolism and quantification of the novel synthetic opioid N-piperidinyl etonitazene (etonitazepipne). Clin Chem Lab Med CCLM 62:1580–1590. 10.1515/cclm-2023-1360
- Berardinelli D, Taoussi O, Daziani G et al (2024b) 3-CMC, 4-CMC, and 4-BMC human metabolic profiling: new major pathways to document consumption of methcathinone analogues? AAPS J 26:70. 10.1208/s12248-024-00940-8
- Bogaards JJP, Bertrand M, Jackson P et al (2000) Determining the best animal model for human cytochrome P450 activities: a comparison of mouse, rat, rabbit, dog, micropig, monkey and man. Xenobiotica 30:1131–1152. 10.1080/00498250010021684
- Boyce M, Favela KA, Bonzo JA et al (2023) Identifying xenobiotic metabolites with in silico prediction tools and LCMS suspect screening analysis. Front Toxicol 5:1051483. 10.3389/ftox.2023.1051483
- Brunetti P, Lo Faro AF, Di Trana A et al (2023) β′-phenylfentanyl metabolism in primary human hepatocyte incubations: identification of potential biomarkers of exposure in clinical and forensic toxicology. J Anal Toxicol 46:e207–e217. 10.1093/jat/bkac065
- Carlier J, Diao X, Wohlfarth A et al (2017) In vitro metabolite profiling of ADB-FUBINACA, a new synthetic cannabinoid. CN. 10.2174/1570159X15666161108123419
- Dalgaard L (2015) Comparison of minipig, dog, monkey and human drug metabolism and disposition. J Pharmacol Toxicol Methods 74:80–92. 10.1016/j.vascn.2014.12.005
- Di Trana A, Brunetti P, Giorgetti R et al (2021) In silico prediction, LC-HRMS/MS analysis, and targeted/untargeted data-mining workflow for the profiling of phenylfentanyl in vitro metabolites. Talanta 235:122740. 10.1016/j.talanta.2021.122740
- Djoumbou-Feunang Y, Fiamoncini J, Gil-de-la-Fuente A et al (2019) BioTransformer: a comprehensive computational tool for small molecule metabolism prediction and metabolite identification. J Cheminform 11:2. 10.1186/s13321-018-0324-5
- Du Q-S, Huang R-B, Chou K-C (2008) Recent advances in QSAR and their applications in predicting the activities of chemical molecules, peptides and proteins for drug design. CPPS 9:248–259. 10.2174/138920308784534005
- Fraser AD, Isner AF, Bryan W (1993) Urinary screening for adinazolam and its major metabolites by the Emit® d.a.u:™ and FPIA benzodiazepine assays with confirmation by HPLC. J Anal Toxicol 17:427–431. 10.1093/jat/17.7.427
- Gerets HHJ, Tilmant K, Gerin B et al (2012) Characterization of primary human hepatocytes, HepG2 cells, and HepaRG cells at the mRNA level and CYP activity in response to inducers and their predictivity for the detection of human hepatotoxins. Cell Biol Toxicol 28:69–87. 10.1007/s10565-011-9208-4
- Gertrudes JC, Maltarollo VG, Silva RA et al (2012) Machine learning techniques and drug design. CMC 19:4289–4297. 10.2174/092986712802884259
- Gicquel T, Pelletier R, Richeval C et al (2021) Metabolite elucidation of 2-fluoro-deschloroketamine (2F-DCK) using molecular networking across three complementary in vitro and in vivo models. Drug Test Anal. 10.1002/dta.3162
- Gicquel T, Pelletier R, Bourdais A et al (2024) Interest of molecular networking in fundamental, clinical and forensic toxicology: a state-of-the-art review. TrAC Trends Analyt Chem. 10.1016/j.trac.2024.117547
- Goncalves R, Pelletier R, Couette A et al (2022) Suitability of high-resolution mass spectrometry in analytical toxicology: focus on drugs of abuse. Toxicol Analyt Clin 34:29–41. 10.1016/j.toxac.2021.11.006
- Guengerich FP (1997) Comparisons of catalytic selectivity of cytochrome P450 subfamily enzymes from different species. Chem Biol Interact 106:161–182. 10.1016/S0009-2797(97)00068-9
- Kazmi SR, Jun R, Yu M-S et al (2019) In silico approaches and tools for the prediction of drug metabolism and fate: a review. Comput Biol Med 106:54–64. 10.1016/j.compbiomed.2019.01.008
- Kirchmair J, Göller AH, Lang D et al (2015) Predicting drug metabolism: experiment and/or computation? Nat Rev Drug Discov 14:387–404. 10.1038/nrd4581
- Lin JH (1995) Species similarities and differences in pharmacokinetics. Drug Metab Dispos 23:1008–1021
- Lin JH (1998) Applications and limitations of interspecies scaling and in vitro extrapolation in pharmacokinetics. Drug Metab Dispos 26:1202–1212
- Lindbom K, Norman C, Baginski S et al (2024) Human metabolism of the semi-synthetic cannabinoids hexahydrocannabinol, hexahydrocannabiphorol and their acetates using hepatocytes and urine samples. Drug Test Anal. 10.1002/dta.3740
- Pelletier R, Gicquel T, Simoes Eugenio M et al (2022a) A transversal approach combining in silico, in vitro and in vivo models to describe the metabolism of the receptor interacting protein 1 kinase inhibitor sibiriline. Pharmaceutics 14:2665. 10.3390/pharmaceutics14122665
- Pelletier R, Le Daré B, Ferron P-J et al (2022b) Use of innovative, cross-disciplinary in vitro, in silico and in vivo approaches to characterize the metabolism of chloro-alpha-pyrrolidinovalerophenone (4-Cl-PVP). Arch Toxicol. 10.1007/s00204-022-03427-7
- Pelletier R, Le Daré B, Le Bouëdec D et al (2023) Identification, synthesis and quantification of eutylone consumption markers in a chemsex context. Arch Toxicol. 10.1007/s00204-023-03615-z
- Pelletier R, Bourdais A, Fabresse N et al (2024) In silico and in vitro metabolism studies of the new synthetic opiate AP-237 (bucinnazine) using bioinformatics tools. Arch Toxicol 98:165–179. 10.1007/s00204-023-03617-x
- Ridder L, Wagener M (2008) SyGMa: combining expert knowledge and empirical scoring in the prediction of metabolites. ChemMedChem 3:821–832. 10.1002/cmdc.200700312
- Salgueiro-Gonzalez N, Béen F, Bijlsma L et al (2024) Influent wastewater analysis to investigate emerging trends of new psychoactive substances use in Europe. Water Res 254:121390. 10.1016/j.watres.2024.121390
- Santos IC, Maia D, Dinis-Oliveira RJ, Barbosa DJ (2024) New psychoactive substances: health and legal challenges. Psychoactives 3:285–302. 10.3390/psychoactives3020018
- Tyzack JD, Kirchmair J (2019) Computational methods and tools to predict cytochrome P450 metabolism for drug discovery. Chem Biol Drug des 93:377–386. 10.1111/cbdd.13445
- UNODC (2023) Early warning advisory (EWA) on new psychoactive substances (NPS)
- Verougstraete N, Verhaeghe A, Germonpré J et al (2023) Identification of etazene (etodesnitazene) metabolites in human urine by LC-HRMS. Drug Test Anal 15:235–239. 10.1002/dta.3377
- Zhu K, Huang M, Wang Y et al (2024) MetaPredictor: in silico prediction of drug metabolites based on deep language models with prompt engineering. Brief Bioinform. 10.1093/bib/bbae374