Interpretable machine learning model to detect chemically adulterated urine samples analyzed by high resolution mass spectrometry

Machine learning-based clinical decision support using laboratory data

Artificial intelligence (AI) and machine learning (ML) are becoming vital in laboratory medicine and the broader context of healthcare. In this review article, we summarized the development of ML models and how they contribute to clinical laboratory workflow and improve patient outcomes. The process of ML model development involves data collection, data cleansing, feature engineering, model development, and optimization. These models, once finalized, are subjected to thorough performance assessments and validations. Recently, due to the complexity inherent in model development, automated ML tools were also introduced to streamline the process, enabling non-experts to create models. Clinical Decision Support Systems (CDSS) use ML techniques on large datasets to aid healthcare professionals in test result interpretation. They are revolutionizing laboratory medicine, enabling labs to work more efficiently with less human supervision across pre-analytical, analytical, and post-analytical phases. Despite contributions of the ML tools at all analytical phases, their integration presents challenges like potential model uncertainties, black-box algorithms, and deskilling of professionals. Additionally, acquiring diverse datasets is hard, and models' complexity can limit clinical use. In conclusion, ML-based CDSS in healthcare can greatly enhance clinical decision-making. However, successful adoption demands collaboration among professionals and stakeholders, utilizing hybrid intelligence, external validation, and performance assessments.

Characterizing lignins from various sources and treatment processes after optimized sample preparation techniques and analysis via ESI-HRMS and custom mass defect software tools

Sample preparation of complex, natural mixtures such as lignin prior to mass spectrometry analysis, however minimal, is a critical step in ensuring accurate and interference-free results. Modern shotgun-MS techniques, where samples are directly injected into a high-resolution mass spectrometer (HRMS) with no prior separation, usually still require basic sample pretreatment such as filtration and appropriate solvents for full dissolution and compatibility with atmospheric pressure ionization interfaces. In this study, sample preparation protocols have been established for a unique sample set consisting of a wide variety of degraded lignin samples from numerous sources and treatment processes. The samples were analyzed via electrospray (ESI)-HRMS in negative and positive ionization modes. The resulting information-rich HRMS datasets were then transformed into the mass defect space with custom R scripts as well as the open-source Constellation software as an effective way to visualize changes between the samples due to the sample preparation and ionization conditions as well as a starting point for comprehensive characterization of these varied sample sets. Optimized conditions for the four investigated lignins are proposed for ESI-HRMS analysis for the first time, giving an excellent starting point for future studies seeking to better characterize and understand these complex mixtures.

Advances in testing for sample manipulation in clinical and forensic toxicology - Part A: urine samples

In many countries, adherence testing is used to monitor consumption behavior or to prove abstinence. Urine and hair are most commonly used, although other biological fluids are available. Positive test results are usually associated with serious legal or economic consequences. Therefore, various sample manipulation and adulteration strategies are used to circumvent such a positive result. In these critical review articles on sample adulteration of urine (part A) and hair samples (part B) in the context of clinical and forensic toxicology, recent trends and strategies to improve sample adulteration and manipulation testing published in the past 10 years are described and discussed. Typical manipulation and adulteration strategies include undercutting the limits of detection/cut-off by dilution, substitution, and adulteration. New or alternative strategies for detecting sample manipulation attempts can be generally divided into improved detection of established urine validity markers and direct and indirect techniques or approaches to screening for new adulteration markers. In this part A of the review article, we focused on urine samples, where the focus in recent years has been on new (in)direct substitution markers, particularly for synthetic (fake) urine. Despite various and promising advances in detecting manipulation, it remains a challenge in clinical and forensic toxicology, and simple, reliable, specific, and objective markers/techniques are still lacking, for example, for synthetic urine.

New algorithms demonstrate untargeted detection of chemically meaningful changing units and formula assignment for HRMS data of polymeric mixtures in the open-source constellation web application

The field of high-resolution mass spectrometry (HRMS) and ancillary hyphenated techniques comprise a rapidly expanding and evolving area. As popularity of HRMS instruments grows, there is a concurrent need for tools and solutions to simplify and automate the processing of the large and complex datasets that result from these analyses. Constellation is one such of these tools, developed by our group over the last two years to perform unsupervised trend detection for repeating, polymeric units in HRMS data of complex mixtures such as natural organic matter, oil, or lignin. In this work, we develop two new unsupervised algorithms for finding chemically-meaningful changing units in HRMS data, and incorporate a molecular-formula-finding algorithm from the open-source CoreMS software package, both demonstrated here in the Constellation software environment. These algorithms are evaluated on a collection of open-source HRMS datasets containing polymeric analytes (PEG 400 and NIST standard reference material 1950, both metabolites in human plasma, as well as a swab extract containing polymers), and are able to successfully identify all known changing units in the data, including assigning the correct formulas. Through these new developments, we are excited to add to a growing body of open-source software specialized in extracting useful information from complex datasets without the high costs, technical knowledge, and processor-demand typically associated with such tools.