Interpretable machine learning model to detect chemically adulterated urine samples analyzed by high resolution mass spectrometry
Open accessThe DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index.
Abstract
Objectives: Urine sample manipulation including substitution, dilution, and chemical adulteration is a continuing challenge for workplace drug testing, abstinence control, and doping control laboratories. The simultaneous detection of sample manipulation and prohibited drugs within one single analytical measurement would be highly advantageous. Machine learning algorithms are able to learn from existing datasets and predict outcomes of new data, which are unknown to the model. Methods: Authentic human urine samples were treated with pyridinium chlorochromate, potassium nitrite, hydrogen peroxide, iodine, sodium hypochlorite, and water as control. In total, 702 samples, measured with liquid chromatography coupled to quadrupole time-of-flight mass spectrometry, were used. After retention time alignment within Progenesis QI, an artificial neural network was trained with 500 samples, each featuring 33,448 values. The feature importance was analyzed with the local interpretable model-agnostic explanations approach. Results: Following 10-fold cross-validation, the mean sensitivity, specificity, positive predictive value, and negative predictive value was 88.9, 92.0, 91.9, and 89.2%, respectively. A diverse test set (n=202) containing treated and untreated urine samples could be correctly classified with an accuracy of 95.4%. In addition, 14 important features and four potential biomarkers were extracted. Conclusions: With interpretable retention time aligned liquid chromatography high-resolution mass spectrometry data, a reliable machine learning model could be established that rapidly uncovers chemical urine manipulation. The incorporation of our model into routine clinical or forensic analysis allows simultaneous LC-MS analysis and sample integrity testing in one run, thus revolutionizing this field of drug testing.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“The ANN model employed in the study exhibited a noteworthy accuracy of 95.4 %.”
From Machine learning-based clinical decision support using laboratory data (2024) · cites this paper as [30]
Citations
Cites
"Why Should I Trust You?": Explaining the Predictions of Any Classifier
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939778 · Cited by 17,555
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z., Citro, C., Corrado, G. S., Davis, A., Dean, J., Devin, M., et al. (2016) · arXiv · doi:10.48550/arxiv.1603.04467 · Cited by 9,408
Building Predictive Models in R Using the caret Package
Kuhn, M. (2008) · Journal of Statistical Software · doi:10.18637/jss.v028.i05 · Cited by 8,863
Smith, C. A., Want, E. J., O'Maille, G., Abagyan, R., & Siuzdak, G. (2006) · Analytical Chemistry · doi:10.1021/ac051437y · Cited by 5,262
MetaboAnalyst 4.0: towards more transparent and integrative metabolomics analysis
Chong, J., Soufan, O., Li, C., Caraus, I., Li, S., Bourque, G., Wishart, D. S., & Xia, J. (2018) · Nucleic Acids Research · doi:10.1093/nar/gky310 · Cited by 3,629
MS-DIAL: data-independent MS/MS deconvolution for comprehensive metabolome analysis
Tsugawa, H., Cajka, T., Kind, T., Ma, Y., Higgins, B., Ikeda, K., Kanazawa, M., VanderGheynst, J., Fiehn, O., & Arita, M. (2015) · Nature Methods · doi:10.1038/nmeth.3393 · Cited by 3,397
An Introductory Review of Deep Learning for Prediction Models With Big Data
Emmert-Streib, F., Yang, Z., Feng, H., Tripathi, S., & Dehmer, M. (2020) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2020.00004 · Cited by 604
Jaffee, W. B., Trucco, E. M., Levy, S., & Weiss, R. D. (2007) · Journal of Substance Abuse Treatment · doi:10.1016/j.jsat.2006.11.008 · Cited by 135
Adulteration of Urine by “Urine Luck”
Wu, A. H. B., Bristol, B., Sexton, K., Cassella, G., Holtman, V., & Hill, D. W. (1999) · Clinical Chemistry · doi:10.1093/clinchem/45.7.1051 · Cited by 95
The Effects of Adulterants and Selected Ingested Compounds on Drugs-of-Abuse Testing in Urine
Dasgupta, A. (2007) · American Journal of Clinical Pathology · doi:10.1309/fqy06f8xktqpm149 · Cited by 92
Research works printed but not linked: 1
- Scholer, A. The effect of urine manipulation on substance abuse testing. ToxChem und KrimTech 2004;71:127–37.
Cited by
Machine learning-based clinical decision support using laboratory data
Çubukçu, H. C., Topcu, D. İ., & Yenice, S. (2024) · Clinical Chemistry and Laboratory Medicine (CCLM) · doi:10.1515/cclm-2023-1037 · Cited by 66
“In the domain of urine sample analysis, Streun et al. conducted a study utilizing ML techniques, speci fi cally (ANN), to detect instances of chemical manipulation [30].”
BackgroundMethodInfluentialWissenbach, D. K., & Steuer, A. E. (2023) · Analytical and Bioanalytical Chemistry · doi:10.1007/s00216-023-04711-w · Cited by 33
“evaluated machine learning, specifically an artificial neural network (ANN), to classify chemically adulterated urine samples based on their LC-HRMS acquisition data [59].”
BackgroundSelf-citationZeng, T., Liang, Y., Dai, Q., Tian, J., Chen, J., Lei, B., Yang, Z., & Cai, Z. (2022) · Chinese Chemical Letters · doi:10.1016/j.cclet.2022.03.020 · Cited by 26
Analytical approaches for food authentication using LIBS fingerprinting
Shin, S., Wu, X., Patsekin, V., Doh, I., Bae, E., Robinson, J. P., & Rajwa, B. P. (2023) · Spectrochimica Acta Part B Atomic Spectroscopy · doi:10.1016/j.sab.2023.106693 · Cited by 25
Steuer, A. E., Brockbals, L., & Kræmer, T. (2021) · Wiley Interdisciplinary Reviews Forensic Science · doi:10.1002/wfs2.1449 · Cited by 24
“Recent developments more and more employed the application of machine learning methods also occasionally in CT and FT applications (Olesti et al., 2019; Streun et al., 2021).”
BackgroundResultSelf-citationStreun, G. L., Steuer, A. E., Poetzsch, S. N., Ebert, L. C., Dobay, A., & Kræmer, T. (2022) · Clinical Chemistry · doi:10.1093/clinchem/hvac045 · Cited by 17
Self-citationThe Application of Machine Learning in Doping Detection
Yang, Q., Xu, W., Sun, X., Chen, Q., & Niu, B. (2024) · Journal of Chemical Information and Modeling · doi:10.1021/acs.jcim.4c01234 · Cited by 11
An Optimized Data and Model Centric Approach for Multi-Class Automated Urine Sediment Classification
Akhtar, S., Hanif, M. I., Rashid, A., Aurangzeb, K., Khan, E. A., Saraoğlu, H. M., & Javed, K. (2024) · IEEE Access · doi:10.1109/access.2024.3385864 · Cited by 9
“Regrettably, conventional microscopic examination-based methods for urine sediment analysis suffer from inefficiency and subjectivity, rendering them error-prone and time-consuming [6], [7].”
BackgroundLetourneau, D. R., Marzullo, B. P., Αλεξανδρίδου, Α., Barrow, M. P., O’Connor, P. B., & Volmer, D. A. (2023) · Analytical and Bioanalytical Chemistry · doi:10.1007/s00216-023-04942-x · Cited by 9
Cost-Effective Drug Testing: Analytical Methodology, Best Practices, and Clinical Utility
Chenoweth, J., LaValley, C., Eberhardt, C. S., Colby, D. K., Albertson, T. E., & Tran, N. K. (2025) · The Journal of Applied Laboratory Medicine · doi:10.1093/jalm/jfae118 · Cited by 8
The role of AI in pre-analytical phase – use cases
Çubukçu, H. C. (2025) · Clinical Chemistry and Laboratory Medicine (CCLM) · doi:10.1515/cclm-2025-1220 · Cited by 6
“…coupled to high-resolution mass spectrometry data from 702 authentic human urine samples treated with fi ve di ff erent chemical adulterants: pyridinium chlorochromate, potassium nitrite, hydrogen peroxide, iodine, and sodium hypo-chlorite, with water serving as the control condition [30].”
InfluentialGrafinger, K. E., Weinmann, W., Pasin, D., Gréen, H., Stove, C. P., Schöning, V., & Hammann, F. (2026) · Forensic Science International · doi:10.1016/j.forsciint.2026.112883 · Cited by 8
Koujan, A. A. (2024) · Journal of Advances in Medicine and Medical Research · doi:10.9734/jammr/2024/v36i85549 · Cited by 1
Dasgupta, A. (2025) · Therapeutic Drug Monitoring · doi:10.1097/ftd.0000000000001360 · Cited by 1
Yasadışı Madde Analizlerinde Preanalitik Evre
Aksun, S., Bağcı, B., Demir, L., Karalar, F. O., & Aksun, M. (2022) · Journal of Turkish Clinical Biochemistry · doi:10.56615/tkbd.2022.18 · Cited by 1
Letourneau, D. R., August, D. D., & Volmer, D. A. (2023) · Journal of Cheminformatics · doi:10.1186/s13321-023-00680-5 · Cited by 1
“[21] When performing high-resolution mass spectrometry on complex, heterogeneous samples such as those containing large, polymeric species (e.g, natural organic matter, oil samples, lignin), there are often patterns within the resulting spectra corresponding to the gain or loss of the repeating…”
BackgroundInternet-Based Products for Beating Urine Drug Tests
Johnson‐Davis, K. L. (2026) · Springer eBooks · doi:10.1007/978-3-032-15015-8_6
Dasgupta, A. (2026) · Springer eBooks · doi:10.1007/978-3-032-15015-8_8
Artificial intelligence: some fundamentals to know its place and potential in forensic toxicology
Desharnais, B., & Devincenzi, T. (2026) · Journal of Analytical Toxicology · doi:10.1093/jat/bkag063
Yurekli, U., & Beğinoğlu, Ö. (2026) · Metabolites · doi:10.3390/metabo16090651
“Advanced mass spectrometric techniques have recently shown promising performance in the identification of chemically adulterated urine specimens and in improving the reliability of toxicological interpretation [14].”
Metadata
- Authors
- Streun, G. L., Steuer, A. E., Ebert, L. C., Dobay, A., & Kraemer, T.
- Affiliations
- University of Zurich (CH)
- Department of Forensic Pharmacology and Toxicology , Zurich Institute of Forensic Medicine, University of Zurich , Zurich , Switzerland
- Venue
- Clinical Chemistry and Laboratory Medicine (CCLM)
- Publisher
- Walter de Gruyter GmbH
- Published
- 2021-03-22
- Volume
- 59
- Issue
- 8
- Pages
- 1392-1399
- ISSN
- 1434-6621 (ISSN-L)
- 1437-4331
- OpenAlex
- W3139333015
- Cited by
- 18
- References
- 33
- Citations by year
- 2026: 5 · 2025: 3 · 2024: 2 · 2023: 5 · 2022: 2 · 2021: 1
- FWCI
- 1.63
- Type
- journal article
- Language
- en
- Open access
- green
- www.zora.uzh.ch
- Licence
- other-oa
- Locations
- Clinical Chemistry and Laboratory Medicine (CCLM) (publishedVersion)
- PubMed (publishedVersion)
- Zurich Open Repository and Archive (University of Zurich) (acceptedVersion, other-oa)
- Zurich Open Repository and Archive (University of Zurich) (submittedVersion, other-oa)
- Zurich Open Repository and Archive (University of Zurich)
- Funders
- Emma Louise Kessler Foundation
- Topics
- Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Toxicology › Forensic Toxicology and Drug Analysis
- Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Metabolomics and Mass Spectrometry Studies
- Social Sciences › Sociology and Political Science › Doping in Sports
- Keywords
- urine adulteration, urine dilution, artificial neural networks, model interpretability, lime, feature importance, biomarkers, LC-MS data, workplace drug testing, doping control
- MeSH terms
- Machine Learning, Chromatography, Liquid, Pharmaceutical Preparations, Humans, Mass Spectrometry, Substance Abuse Detection
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.