Analytical Chemistry

Sphingolipids are diverse lipids with sphingobases and N-acyl fatty acids as the hydrophobic moieties. While the importance of the in-depth elucidation of hydrophobic structures is widely recognized in lipid biology, mass spectrometry-based annotation of ceramides in the commonly used protonated form is often hindered by in-source dehydration during electrospray ionization in the heated state and variable water losses in the product ion spectrum. In this study, we investigated the sodium ion form and its product ions in ceramides with the use of electron-activated dissociation tandem mass spectrometry (EAD MS/MS) in addition to collision-induced dissociation to facilitate indepth structural elucidation. While dehydrated ions from the protonated form were frequently observed, the sodium adduct ions remained stable because of their higher activation energy compared with the protonated form, which was validated using quantum chemical calculations. Using the three adduct forms under optimized conditions increased confidence in annotating the ceramide peaks through retention-time matching. Furthermore, EAD MS/MS of the sodium adduct ions facilitated the positional determination of double bonds and hydroxyl groups in the ceramide hydrophobic moiety. Our approach is showcased by the annotation of phytoceramides with N-acyl 2- and 3-hydroxyl groups in mouse feces and ceramides with N-acyl n-6 very long-chain polyunsaturated 2-hydroxy fatty acids in mouse testis.

Mass spectrometry (MS) is a powerful technique for characterizing modified RNA as it directly sequences and quantifies all mass-altering modifications simultaneously. However, the physicochemical properties of RNA result in poor ionization efficiencies during electrospray ionization, presenting a major barrier to sensitive MS measurements necessary for low abundance RNA samples and RNAs with low modification stoichiometries. Here, we report a ligation-based approach to increase ionization efficiencies of RNA oligonucleotides. We show that short ([~]5 nt), chemically modified DNA oligonucleotides can be enzymatically ligated to RNA to serve as MS signal enhancers. Among a series of signal enhancers appended with various alkyl and alkylimidazolium functional groups, we found that decyl-functionalized derivatives improved MS sensitivity by [~]15-fold compared to unlabeled oligonucleotide. When ligated to RNA standards, the decyl-modified signal enhancer increased MS signals 2-4-fold with the additional benefit of improved retention during liquid chromatography (LC) separations without ion pairing agents. To apply the ligation-based approach to RNase T1 digests of longer RNAs, a multi-step enzymatic approach was optimized to maximize ligation efficiencies. We then ligated signal enhancers to a yeast transfer RNA (tRNA) digest and observed increased MS signals for numerous sequence-informative digestion products. Importantly, the sequences of RNA oligonucleotides ligated to signal enhancers were readily determined by tandem mass spectrometry with collision-induced dissociation. This ligation-based strategy for enhancing LC-MS/MS characterization of RNA creates opportunities to measure low abundance RNA samples and their modifications.

High-resolution binding site mapping is important for in-depth activity assessment of new therapeutics including AI-designed antibodies. However, complex protein targets such as glycosylated antigens are challenging for many methods including crystallography. PD1 is a highly glycosylated antigen, and with the traditional HDX-MS method, only 51% sequence coverage could be obtained with multiple epitope residues undetected for Pembrolizumab. By implementing glyco-peptide detection, subzero temperature LC-MS and electron based MSMS fragmentation, the new HDX FineMapping methodology enabled 100% sequence coverage and complete epitope characterization for the Pembrolizumab-PD1 system, with amino acid level resolution. Furthermore, HDX FineMapping detects binding epitopes directly in solution, without any mutation or modification to either the antigen or the antibody. The amino acid level resolution combined with low cost, minimal sample consumption, fast turnaround time, and no need of mutant library or crystallization makes it a competitive methodology for binding mode validation of AI-designed therapeutics.

Microbes and bile acids are tightly intertwined, especially in the gut. While the liver produces primary bile acids from cholesterol, gut bacteria transform these into diverse secondary forms which act as powerful signaling molecules, influencing host metabolism and immune function. Since bile acid changes are increasingly linked to health and disease, their accurate measurement in the gut and circulation is essential. Analytical evaluations, however, remain challenging as many bile acids co-elute in liquid chromatography (LC), share identical precursor masses in mass spectrometry (MS), and produce similar tandem mass spectrometry (MS/MS) spectra. As a result, conventional LC-MS/MS workflows struggle to differentiate bile acids, motivating the addition of orthogonal separations such as ion mobility spectrometry (IMS). Here, we assess optimal bile acid extraction parameters for stool, serum, and plasma; compare LC conditions; and assess electrospray ionization performance across polarities. Additionally, we created a multidimensional reference library containing LC retention times, IMS collision cross section values, and accurate precursor masses for 280 unique bile acids (264 endogenous and 16 deuterium-labeled species) including unconjugated, host-conjugated, and microbially conjugated bile acids. This multidimensional library empowers bile acid identification in complex samples and enables a more comprehensive exploration of their biological roles and disease associations.

Machine learning (ML)-assisted Raman spectroscopy has become a powerful analytical tool for the classification and identification of analytes; however, technical challenges impacting its detection accuracy have not been investigated. This study explores experimental factors affecting classification performance. Among the evaluated ML models, ML algorithms show minimal impacts on classification accuracy. Instead, experimental factors, including spectral similarity between tested samples and the data quality, dominate detection performance. Increases in spectral noises and spectral similarity significantly reduce classification accuracy. In well-controlled samples with low experimental noise, ML-assisted Raman spectroscopy can discriminate lipid mixtures with a composition difference of 1.85 mol%. To assess the effect of biological heterogeneity, we analyzed single-cell Raman spectra from Saccharomyces cerevisiae strains carrying single, double, or triple gene mutations. Intrinsic cell-to-cell variability introduced substantial spectral differences, severely reducing the accuracy of multiclass classification of these genetically similar strains at the single-cell level. Averaging Raman spectra across multiple cells improved classification accuracy by reducing this spectral variability. We also assess the effectiveness of transfer learning across different Raman spectrometers, specifically by applying a ML model trained on one instrument to another Raman spectrometer. Transfer learning can be improved with proper instrument calibration, highlighting the importance of instrument standardization. Overall, our results demonstrate that data quality and spectral similarity are the primary bottlenecks in ML-assisted Raman spectroscopy. Careful attention to sample preparation, data acquisition, measurement conditions, and instrument calibration is critical to achieving robust and reliable classification performance.

Polar metabolites, including amino acids, nucleotides, phosphorylated metabolites, and central carbon intermediates, drive essential physiological processes but remain difficult to measure by high-throughput, reversed-phase LC-MS due to poor retention on conventional stationary phases. We developed a 3-minute reversed-phase LC-MS method and benchmarked T3-type C18 and pentafluorophenyl (PFP) chemistries for profiling 123 polar metabolites across acidic and mildly acidic conditions. The T3 phase operated under mildly acidic conditions provided the best overall performance, achieving the highest coverage, robust retention-time stability, and improved detection of phosphorylated metabolites. To strengthen compound annotation under ultra-short gradients, we combined the method with iterative data-dependent MS/MS, acquiring spectra for 86 of 123 metabolite mixture compounds without extending runtime. Retention times and peak shapes remained stable over 480 consecutive injections (mean CV of 1.7%) in Escherichia coli extract. Together, these results define a rapid, scalable workflow for profiling of polar and phosphorylated metabolites on standard instrumentation for high-throughput biological studies.

Untargeted proteomics enables quantitative determination of host cell proteins (HCPs) in biotherapeutics, yet no workflow has been validated under ICH Q2(R2) for regulated quality control. We report a prospective validation of label-free untargeted proteomics for HCP quantification using a total-error (TE) approach. A stable isotope-labeled whole-proteome standard was spiked into NISTmAb at seven levels (20-80 ng). Four independent assays (198 injections) supported hierarchical replication and one-way random-effects ANOVA variance decomposition with Welch-Satterthwaite adjustment. Dual entrapment analysis demonstrated empirical peptide-level false discovery proportions below 1% at q = 0.01. Deterministic parsimony inference ensured invariant protein-group definition. Weighted least-squares regression (R{superscript 2} = 0.993) identified stable proportional compression with recoveries of 81-85%. Repeatability dominated the variance structure (median CV 2.7%); intermediate precision total SD ranged from 0.69% to 3.81% over the validated range. Accuracy profiles integrating empirical bias with a log- log variance model showed 95% {beta}-expectation and 95/95 content tolerance intervals fully contained within {+/-}30%, with a lower limit of quantification (LLOQ) of 20 ng. Abundance-stratified TE analysis revealed concentration-dependent calibration heterogeneity masked by aggregate-level estimation; stratum-specific {beta}-expectation intervals within {+/-}35% defined an abundance-aware LLOQ of 3.6 ppm (P95 = 3.87 ppm). Robustness under independent search software (FragPipe, CCC = 0.998, LoA {+/-}9%) and cross-platform acquisition (Astral, CCC = 0.980, LoA {+/-}18%) remained within predefined {+/-}30% agreement limits. System suitability criteria were derived empirically from validation performance. This is the first prospective ICH Q2(R2)-aligned validation of untargeted proteomics for HCP quantification, with a statistical framework applicable to other high-dimensional analytical methods requiring regulatory qualification. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=113 SRC="FIGDIR/small/710150v1_ufig1.gif" ALT="Figure 1"> View larger version (29K): org.highwire.dtl.DTLVardef@1f5331aorg.highwire.dtl.DTLVardef@ee2234org.highwire.dtl.DTLVardef@798eaorg.highwire.dtl.DTLVardef@c84034_HPS_FORMAT_FIGEXP M_FIG C_FIG

Capillary zone electrophoresis (CZE)-mass spectrometry (MS) has been proposed as a powerful analytical tool for bottom-up, top-down, and native proteomics (multi-level proteomics) decades ago to analyze complex biological samples at the levels of peptides (bottom-up), proteoforms (top-down), and complexoforms (native). However, its broad adoption has been impeded by the limited robustness and reproducibility. Here, we present multi-level proteomics data from nearly 170 CZE-MS runs ([~]170 hours of instrument time), demonstrating qualitatively (i.e., the number of identified peptides and proteoforms, the number of detected complexoforms, and their migration time) and quantitatively (i.e., peptide, proteoform, and complexoform intensity) reproducible measurement of complex samples with varying levels of complexity, i.e., Escherichia coli cells, HeLa cells, and human plasma. CZE-MS-based native proteomics enabled the detection of hundreds of complexoforms up to 800 kDa from the complex systems via consuming only nanograms of protein material. The results indicate that CZE-MS is sensitive and reproducible enough for broad adoption for multi-level proteomics-based biomedical research.

Limited proteolysis coupled to mass spectrometry (LiP-MS) is a peptide-centric conformational proteomics approach during which a brief incubation with a non-specific protease (e.g., proteinase K) under native conditions generates structural fingerprints that report on treatment-induced conformational changes, which is followed by a tryptic digest under denaturing conditions allowing to read out these fingerprints 1. In contrast, the recently introduced peptide-centric local stability assay (PELSA) uses a high trypsin-to-substrate ratio under native conditions to release fully tryptic peptides that reflect structural stability upon ligand binding 2. In their paper, Li et al. compared PELSA and LiP-MS across several benchmarks and reported that PELSA exhibited quantitative sensitivity comparable to or exceeding LiP-MS. Notably, PELSA quantified a 21-fold greater rapamycin-induced change for FKBP1A compared to LiP-MS. Because such claims influence method selection for conformational proteomics, we reanalyzed the publicly deposited datasets underlying these comparisons and assessed the experimental and analytical choices that contributed to the reported effect sizes. Our evaluation indicates that the reported 21-fold difference arises from non-matched experimental conditions and undisclosed data imputation, and that conclusions regarding quantitative superiority or biological interpretability should therefore be treated with caution.

Plasma contains diverse bioactive peptides that play crucial roles in maintaining homeostasis and regulating disease responses. However, the presence of peptides derived from high-abundance proteins such as albumin makes comprehensive analysis of native peptides secreted by organs challenging. This study aimed to establish a highly sensitive plasma peptidomic approach by combining data-independent acquisition (DIA) with spectral libraries of plasma and organs. First, peptides were extracted from plasma and eleven organ types using a high-yield peptide extraction method, the differential solubilization method. These peptides were then measured via data-dependent acquisition (DDA) analysis using a timsTOF HT for constructing empirical spectral library. Subsequently, DIA-MS data from plasma samples were measured and analyzed using this spectral library. This strategy achieved identification of, on average, over 5,500 peptides per run, with over 2,000 organ-derived peptides including 19 known bioactive peptides. The novel strategy proposed here enables highly sensitive quantitative analysis of organ-derived peptides in plasma, linking them to their secreting organs. It is expected to substantially contribute not only to the discovery of biomarkers and novel bioactive peptides but also to elucidating the pathophysiology of systemic diseases.

Small extracellular vesicles (sEVs) are membrane-bound particles whose protein, lipid, and metabolite cargo reflects the molecular state of their cells of origin, making them attractive targets for biomarker discovery and therapeutic development. However, comprehensive characterization of sEVs remains challenging due to the extremely limited material available. Here, we present an integrated mass spectrometry-based multi-omics platform for simultaneous characterization of lipids, metabolites, and proteins from a single sEV sample enabled by sequential extraction, maximizing sample utilization. To enhance molecular coverage and analytical depth, the platform combines iterative tandem mass spectrometry for improved small-molecule fragmentation and nano-flow proteomics with data-independent acquisition. We achieved deep and reproducible multi-omic characterization of proteins, lipids, and metabolites using 10 million sEVs. We further demonstrated the compatibility of our multi-omics platform with sEVs isolated from plasma by ultracentrifugation, size-exclusion chromatography with ultrafiltration, and polymer precipitation, revealing purification-dependent differences in molecular profiles associated with tradeoffs in yield and purity of sEVs. By enabling integrated multi-omics from the same sample, this strategy addresses a key challenge in low-input sEV analysis and establishes a robust analytical foundation for synergistic biomarker discovery and therapeutic applications.

Ambient metabolomics techniques such as laser-assisted rapid evaporative ionization mass spectrometry (LA-REIMS) enable fast, preparation-free fingerprinting of biological samples but are inherently limited by spectral congestion in the absence of chromatographic separation. While ion mobility spectrometry provides additional gas-phase separation, maintaining ion transmission under the transient signals characteristic of laser desorption, remains analytically challenging. Here, we define operating conditions for cyclic traveling-wave ion mobility spectrometry (cIMS) that preserve transmission under LA-REIMS duty-cycle constraints and systematically evaluate how cIMS integration reshapes biofluid fingerprints and enhances chemical specificity in chromatography-free metabolomics analysis. Under optimized single-pass conditions, cIMS separation reorganized LA-REIMS spectra into structured mass/mobility feature domains, enabling selective mobility-based filtering of matrix-derived salt cluster ions. This reduced non-biological background contributions by up to 35% of total spectral intensity while preserving over 90% of detected untargeted features. Although cIMS operation introduced a sensitivity penalty relative to time-of-flight-only acquisition, approximately 80% of the total ion current was recovered under optimized conditions. Mobility-resolved data revealed coherent homologous series and class-specific structural trends, particularly for lipids, supporting class-level annotation. Analysis of 101 metabolite and lipid standards covering a broad physicochemical range (logP -5.30 to 19.40) demonstrated comprehensive molecular coverage, high mass accuracy (mean 2.4 ppm), and good agreement with reference CCS values (mean deviation 4.0%), with isomer separation observed for biologically important secondary bile acids in extended separation cycles. Collectively, these results establish LA-REIMS-cIMS as a practical analytical strategy for enhancing chemical specificity and spectral interpretability in support of high-throughput large-scale metabolic fingerprinting. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=147 SRC="FIGDIR/small/709786v1_ufig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@18a2dfdorg.highwire.dtl.DTLVardef@d165d6org.highwire.dtl.DTLVardef@1750291org.highwire.dtl.DTLVardef@fbbce9_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOGraphical abstractC_FLOATNO Ion mobility spectrometry adds an orthogonal gas-phase separation to LA-REIMS, reorganizing complex biofluid spectra into distinct mass-mobility feature bands and improving molecular resolution in rapid ambient ionization metabolomics. C_FIG

This study investigated the effects of different downstream processes for protein isolation on the bulk properties and composition of clover grass protein prototypes (CGPs). A clarified clover grass juice, obtained using membrane filtration (MF), underwent precipitation by acid (AP), heat (HP), or acid+heat (AHP), or underwent ultra- and diafiltration to produce a concentrate (DC) as well as subsequent tryptic hydrolysis of DC (DCH). HP had the highest protein content (p<0.05) and was whiter than other CGPs, although it showed lower aqueous solubility. In contrast, DC showed excellent solubility across a broad pH range. CGPs efficiently decreased oil-water interfacial tension (16-13 mN/m) and displayed viscoelastic and solid-like interfacial behavior. CGPs-stabilized emulsions displayed low physical stability with larger droplets despite high absolute {zeta}-potentials. CGPs were rich in RuBisCO (37-47%) but had varying levels of other proteins. Despite significant protein-level differences, overall protein composition of CGPs was comparable, highlighting that protein state governs bulk functionality more than subtle compositional changes. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=108 SRC="FIGDIR/small/701969v1_ufig1.gif" ALT="Figure 1"> View larger version (25K): org.highwire.dtl.DTLVardef@f37756org.highwire.dtl.DTLVardef@1fb5beorg.highwire.dtl.DTLVardef@1d4efe2org.highwire.dtl.DTLVardef@d11ef8_HPS_FORMAT_FIGEXP M_FIG C_FIG Created with BioRender.com HighlightsO_LIThe effect of different processes on functional properties of CGPs was explored. C_LIO_LIHeat treatment increased protein purity and whiteness at the expense of solubility. C_LIO_LICGPs efficiently reduced O/W interfacial tension but produced unstable emulsions. C_LIO_LICGPs were found rich in RuBisCO (34-47%) using quantitative proteomics. C_LIO_LIProtein state had larger influence on functionality than protein-level composition. C_LI

Stable carbon and nitrogen isotope ratios are widely used in the life sciences to investigate diet, trophic interactions, and metabolic fluxes, but conventional isotope ratio mass spectrometry requires milligram-scale samples, limiting its applicability to small or rare biological specimens. Fourier Transform Isotopic Ratio Mass Spectrometry (FT IsoR MS) enables amino acid-resolved isotope analysis in a proteomics-compatible workflow and has previously been demonstrated at the microgram scale. Here, we assess the lower sample limit of FT IsoR MS by integrating it with single-cell proteomics-style sample preparation. Using human HeLa cells cultured in 13C-glucose-enriched and control media, we show that reliable relative {delta}13C measurements can be obtained from as few as 50 cells, corresponding to <10 ng of total protein, with a precision of approximately {+/-}9{per thousand}. The observed amino acid-specific labeling patterns are metabolically coherent and consistent with bulk measurements, while smaller cell numbers ([&le;]10 cells) do not yield statistically robust results. These findings establish the practical sensitivity threshold of FT IsoR MS at the low-nanogram level and demonstrate its suitability for isotope-resolved analyses of small cell populations, micro-organoids, and other low-input biological samples, thereby extending stable isotope analysis toward single-cell-scale applications.

Cultivated meat production requires robust and validated analytical methods for comprehensive characterization. While transcriptomics-based approaches establish the foundational profile of molecular analysis, proteomics provides additional resolution that further enhances scientific certainty in both product development and safety characterization. However, the industry adoption of proteomics is currently hindered by technical complexity and a critical lack of analytical standardization, which leads to significant workflow-dependent variations in proteome coverage. To address this gap, we investigated the influence of key workflow steps (digestion, cleanup, LC-MS conditions) on the proteome profile of cultivated duck biomass. We compared five bottom-up sample preparation protocols - two traditional in-solution options (urea and SDC-based protocols), two device-based approaches (PreOmics iST and EasyPep kits), and an innovative protocol (SPEED), and demonstrated that device-based protocols offered the highest peptide yield and proteome coverage. However, optimization allowed cost-effective in-solution methods to achieve comparable performance. Specifically, an optimal digestion time of 3 hours at 37{degrees}C and the use of polymer-based desalting columns significantly enhanced protein identification ([~]4500 - 5000 IDs). Moreover, data independent acquisition (DIA) provided deeper proteome coverage than data dependent acquisition (DDA) with higher precision ([~]6500 vs 5000 IDs). The validated Standard Operating Procedures presented here establish a standardized framework for bulk bottom-up proteomics in cultivated meat, facilitating the generation of reliable and comparable data required for robust multi-omics characterization. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=80 SRC="FIGDIR/small/713501v1_ufig1.gif" ALT="Figure 1"> View larger version (32K): org.highwire.dtl.DTLVardef@5b61b8org.highwire.dtl.DTLVardef@16c7e65org.highwire.dtl.DTLVardef@1de21d2org.highwire.dtl.DTLVardef@7e984a_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsO_LIComplexity and non-standardization limit MS-proteomics use in cultivated meat (CM). C_LIO_LICM protein profile varies with sample prep, LC-MS, and data processing pipeline. C_LIO_LIDevice-based and optimized cost-effective protocols offer a high proteome coverage. C_LIO_LIProteomics can complement transcriptomics for a comprehensive CM characterization. C_LIO_LIProposed standardized methods ensure reliable data for future regulatory submissions. C_LI

In this study, we demonstrate that urea enables reliable melting temperature (Tm) determination of hyperthermostable proteins by nano differential scanning fluorimetry (nanoDSF) Under native conditions, Pfu DNA polymerase and its Sso7d-fusion variant showed no detectable unfolding transitions, despite their Tm values falling within the instruments operational range, reflecting their extreme kinetic stability. In the presence of up to 7 M urea, intrinsic tyrosine and tryptophan fluorescence revealed clear unfolding transitions, yielding extrapolated Tm values of 104.8 {+/-} 0.09 {degrees}C for Pfu and 106.8 {+/-} 0.33 {degrees}C for its Sso7d-fusion variant. These results demonstrate that urea-gradient nanoDSF overcomes both instrumental and kinetic limitations, providing a simple and robust method for assessing the thermal stability of (hyper)thermostable proteins. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=59 SRC="FIGDIR/small/717478v1_ufig1.gif" ALT="Figure 1"> View larger version (16K): org.highwire.dtl.DTLVardef@1b8c5e3org.highwire.dtl.DTLVardef@1c7d395org.highwire.dtl.DTLVardef@14093forg.highwire.dtl.DTLVardef@16b2f25_HPS_FORMAT_FIGEXP M_FIG C_FIG

Recent advances in liquid chromatography-mass spectrometry (LC-MS) have accelerated the adoption of high-throughput workflows that deliver deep proteome coverage using minimal sample amounts. This trend is largely driven by clinical and single-cell proteomics, where sensitivity and reproducibility are essential. Here, we extend our previous benchmark dataset (PXD028735) using next-generation LC-MS platforms optimized for rapid proteome analysis. We generated an extensive DDA/DIA dataset using a human-yeast-E. coli hybrid proteome. The proteome sample was distributed across multiple laboratories together with standardized analytical protocols specifying two short LC gradients (5 and 15 min) and low sample input amounts. This dataset includes data acquired on four different platforms, and features new scanning quadrupole-based implementations, extending coverage across different instruments and acquisition strategies. Our comprehensive evaluation highlights how technological advances and reduced LC gradients may affect proteome depth, quantitative precision, and cross-instrument consistency. The release of this benchmark dataset via ProteomeXchange (PXD070049 and PXD071205), allows for the acceleration of cross-platform algorithm development, enhance data mining strategies, and supports standardization of short-gradient, high-throughput LC-MS-based proteomics.

BackgroundConstructing digital twins in healthcare requires biological data sources that are simultaneously informative, dynamic, and practical for routine collection. Dried blood spot (DBS) sampling combined with untargeted metabolomics is well suited to meet these requirements: DBS can be self-collected at home and mailed at ambient temperature, while untargeted LC-MS/MS captures thousands of metabolites reflecting individual physiology, lifestyle, and exposures. We previously demonstrated proof-of-concept individual identification from DBS-derived metabolomic profiles in 277 volunteers (80-92% accuracy). Here, we report a large-scale validation on a substantially expanded cohort. MethodsWe collected 18,288 DBS samples from 1,257 individuals across 134 analytical batches over 15 months. Samples were self-collected at home, mailed via standard postal service, and analyzed by untargeted LC-MS/MS on a high-resolution Orbitrap platform in positive ESI mode. Our classification pipeline comprises batch-aware normalization, supervised feature selection, biological signal filtering, dimensionality reduction, and user-level majority voting across all available samples. This voting reflects the real-world use case: participants contribute multiple self-collected DBS cards over time, taken at different times of day and under varying conditions. We employed GroupKFold cross-validation with group=batch to ensure zero batch leakage between training and testing sets. ResultsIn 10-fold GroupKFold cross-validation (group=batch, zero batch leakage), our pipeline achieved 94.1% user-level identification accuracy (85.5% sample-level). In a fully held-out validation on 17 future batches -- with all feature selection, normalization, and model fitting performed exclusively on training data -- performance was even stronger: 96.1% user-level and 92.6% sample-level across 1,134 classes (chance level: 0.088%). Feature selection stability was confirmed via bootstrap analysis. We identified batch leakage as a critical methodological pitfall for the field: naive random splitting inflated accuracy by sharing 92.8% of test samples (user, batch) pairs with the training set. The top discriminative metabolites span biologically relevant pathways including amino acid metabolism, fatty acid transport, and sphingolipid biosynthesis. ConclusionsUntargeted metabolomics from dried blood spots supports batch-aware, closed-set individual identification in a single-laboratory setting, with potential relevance for longitudinal sample-to-person linkage in future digital twin workflows.

Carbapenemases are major drivers of carbapenem resistance in Gram-negative bacteria and pose a critical threat to last-line antibiotic therapy. Rapid identification of carbapenemase classes is essential for appropriate treatment and epidemiological surveillance; however, current functional methods lack class-level resolution and may yield false-negative results for OXA-48-like enzymes. In this study, we developed and validated an assay based on liquid chromatography-mass spectrometry with trapped ion mobility spectrometry-time-of-flight [LC-MS (timsTOF)] for simultaneous detection and class-level differentiation of five clinically relevant carbapenemases (KPC, NDM, VIM, IMP, and OXA-48-like). The method employs three carbapenem substrates (meropenem, imipenem, and ertapenem). A total of 55 clinical isolates were analyzed using a standardized 2-hour incubation protocol, with a total analysis time of 7 min per sample. Ion mobility enabled unambiguous identification of the OXA-48-specific meropenem-derived {beta}-lactone based on its distinct collisional cross-section (185 [A]{superscript 2} vs. 191 [A]{superscript 2} for intact meropenem), despite identical mass and nearly identical retention time. This marker was detected in all OXA-48-like producers and was absent in all other groups. In contrast, imipenem and ertapenem did not provide comparable discrimination, highlighting the central role of meropenem. Distinct hydrolysis profiles enabled class-level differentiation supported by multivariate analysis. LC-MS (timsTOF) thus enables rapid, sensitive, and specific functional detection of carbapenemases within a single workflow. The ion mobility dimension is critical for accurate identification of OXA-48-like enzymes and supports the potential implementation of this approach in routine clinical microbiology laboratories. ImportanceThis study introduces an ion mobility-enabled LC-MS (timsTOF) approach for functional detection and class-level differentiation of clinically relevant carbapenemases within a single analytical workflow. By leveraging collisional cross-section measurements, the method enables reliable identification of OXA-48-like carbapenemase through detection of a meropenem-derived {beta}-lactone that is indistinguishable by mass alone. This directly addresses a major diagnostic limitation of conventional activity-based assays, which may yield false-negative results for OXA-48-like enzymes. The approach further demonstrates the potential of integrating ion mobility into routine clinical mass spectrometry to enhance specificity beyond traditional mass and retention time measurements. These findings support the development of next-generation diagnostic strategies capable of detecting both known and emerging resistance mechanisms without reliance on predefined targets.

A machine learning approach to molecular formula assignment is crucial for unlocking the full potential of ultra-high resolution mass spectrometry (UHRMS) when analyzing complex mixtures. By combining data-driven models with rigorous benchmarking, the accuracy, consistency, and speed in identifying plausible molecular formulas from vast spectral datasets can be improved. Compared with traditional de novo methods that rely heavily on rule-based heuristics, and manual parameter tuning, machine learning approaches can capture complex patterns in data and adapt more readily to diverse sample types. In this paper, we describe the application of a machine learning methods using the k-nearest neighbors (KNN) algorithm trained on curated chemical formula datasets of UHRMS analysis of dissolved organic matter (DOM) covering the saline river continuum and tropical wet/dry season variability. The influence of the mass accuracy (training set with 0.15-1ppm) was evaluated on a blind test set of DOMs of different geographical origins. A Decision Tree Regressor (DTR) and Random Forest Regressor (RFR) based on mass accuracy (<1ppm) was used. Results from our ML models exhibit 43% more formulas annotated than traditional methods (5796 vs 4047), Model-Synthetic achieved 99.9% assignment rate and annotated/assigned 2x more formulas (8,268 vs 4047). DTR and RFR achieved formula-level accuracies (FA) of 86.5% and 60.4%, respectively. Overall, results show an increase in formula assignment when compared with traditional methods. This ultimately enables more reliable characterization of complex natural and engineered systems, supporting advances in fields such as environmental science, metabolomics, and petroleomics. Furthermore, the novel data set produced for this study is made publicly available, establishing an initial benchmark for molecular formula assignment in UHRMS using machine learning. The dataset and code are publicly available at: https://github.com/pcdslab/dom-formula-assignment-using-ml CCS CONCEPTSComputing methodologies [-&gt;] Machine Learning [-&gt;] Learning paradigms [-&gt;] Supervised Learning