Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2023 Dec 19;95(50):18326-18334.
doi: 10.1021/acs.analchem.3c02413. Epub 2023 Dec 4.

Deep Learning-Enabled MS/MS Spectrum Prediction Facilitates Automated Identification Of Novel Psychoactive Substances

Fei Wang  1   2 , Daniel Pasin  3 , Michael A Skinnider  4   5   6 , Jaanus Liigand  7   8 , Jan-Niklas Kleis  9 , David Brown  10   11 , Eponine Oler  7 , Tanvir Sajed  7 , Vasuk Gautam  7 , Stephen Harrison  10 , Russell Greiner  1   2 , Leonard J Foster  4   12 , Petur Weihe Dalsgaard  3 , David S Wishart  1   7   13   14   15
Affiliations

Deep Learning-Enabled MS/MS Spectrum Prediction Facilitates Automated Identification Of Novel Psychoactive Substances

Fei Wang et al. Anal Chem. .

Abstract

The market for illicit drugs has been reshaped by the emergence of more than 1100 new psychoactive substances (NPS) over the past decade, posing a major challenge to the forensic and toxicological laboratories tasked with detecting and identifying them. Tandem mass spectrometry (MS/MS) is the primary method used to screen for NPS within seized materials or biological samples. The most contemporary workflows necessitate labor-intensive and expensive MS/MS reference standards, which may not be available for recently emerged NPS on the illicit market. Here, we present NPS-MS, a deep learning method capable of accurately predicting the MS/MS spectra of known and hypothesized NPS from their chemical structures alone. NPS-MS is trained by transfer learning from a generic MS/MS prediction model on a large data set of MS/MS spectra. We show that this approach enables a more accurate identification of NPS from experimentally acquired MS/MS spectra than any existing method. We demonstrate the application of NPS-MS to identify a novel derivative of phencyclidine (PCP) within an unknown powder seized in Denmark without the use of any reference standards. We anticipate that NPS-MS will allow forensic laboratories to identify more rapidly both known and newly emerging NPS. NPS-MS is available as a web server at https://nps-ms.ca/, which provides MS/MS spectra prediction capabilities for given NPS compounds. Additionally, it offers MS/MS spectra identification against a vast database comprising approximately 8.7 million predicted NPS compounds from DarkNPS and 24.5 million predicted ESI-QToF-MS/MS spectra for these compounds.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing financial interest.

Figures

Figure 1
Figure 1
Performance on the compound-to-mass spectrum (C2MS) task. The NPS-MS De Novo model is trained from scratch on the NPS training data set, whereas the NPS-MS model is trained via transfer learning from a CFM-ID 4.0 base model that is subsequently fine-tuned on the NPS training data set. Bars display mean scores for each metric with error bars indicating the 95% confidence interval. Left, the overall performance of each model averaged over three different collision energies (10, 20, and 40 eV). Right, the performance of each model on MS/MS spectra collected at each individual collision energy.
Figure 2
Figure 2
Performance on the mass spectrum-to-compound (MS2C) tasks. ID Task A: MS2C identification task on HighResNPS data set (Data Set #1). ID Task B: MS2C identification task on HighResNPS + PubChem data set (Data Set #2). ID Task B. SIRIUS: MS2C identification task was performed on a subset of Data Set #2 to enable comparison with SIRIUS 4. ID Task C: MS2C identification task on HighResNPS + DarkNPS data set (Data Set #3). For each task, shown from left to right are far left, cost-1 score; middle left, CDF of cost-1 to cost-10 score; middle right, Tanimoto coefficient between the highest-ranked candidate and the ground truth structure; far right, negative CDDD distance of the highest-ranked candidate and the ground truth structure.
Figure 3
Figure 3
(a). Identification of the designer fentanyl derivative ocfentanil in an MS2C task. The top-3 candidates predicted by NPS-MS from Data Set #2 based on the spectra of ocfentanil were all structurally related to ocfentanil. The correct structure was ranked as the third-best candidate, and the top-2 candidates were its 3-fluoro (Tc = 0.64) and 4-fluoro (Tc = 0.67) isomers. In contrast, the top-3 candidates given by CFM-ID 4.0 display little resemblance to the correct structure, with Tanimoto coefficients of 0.12, 0.09, and 0.23. (b) Retrospective application of NPS-MS to identify an unknown NPS detected in a seized powder. Left, MS/MS spectra of 3-Cl-PCP predicted by CFM-ID 4.0 at 10, 20, and 40 eV. Middle, MS/MS spectra of 3-Cl-PCP predicted by NPS-MS at 10, 20, and 40 eV. Right, the top-3 compounds identified by NPS-MS and CFM-ID 4.0 in a MS2C identification task when searching the experimentally acquired 3-Cl-PCP spectra against a database of novel chemical structures anticipated by DarkNPS.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

References

    1. Hill S. L.; Thomas S. H. L. Clinical Toxicology of Newer Recreational Drugs. Clin. Toxicol. 2011, 49 (8), 705–719. 10.3109/15563650.2011.615318. - DOI - PubMed
    1. Baumann M. H.; Volkow N. D. Abuse of New Psychoactive Substances: Threats and Solutions. Neuropsychopharmacology 2016, 41 (3), 663–665. 10.1038/npp.2015.260. - DOI - PMC - PubMed
    1. Varì M. R.; Mannocchi G.; Tittarelli R.; Campanozzi L. L.; Nittari G.; Feola A.; Umani Ronchi F.; Ricci G. New Psychoactive Substances: Evolution in the Exchange of Information and Innovative Legal Responses in the European Union. Int. J. Environ. Res. Public Health 2020, 17 (22), 8704.10.3390/ijerph17228704. - DOI - PMC - PubMed
    1. Reuter P.; Pardo B. New Psychoactive Substances: Are There Any Good Options for Regulating New Psychoactive Substances?. Int. J. Drug Policy 2017, 40, 117–122. 10.1016/j.drugpo.2016.10.020. - DOI - PubMed
    1. Grafinger K. E.; Bernhard W.; Weinmann W. Scheduling of New Psychoactive Substance the Swiss Way: A Review and Critical Analysis. Sci. Justice 2019, 59 (4), 459–466. 10.1016/j.scijus.2019.03.005. - DOI - PubMed

Publication types

MeSH terms