The turn of the 19th and 20th centuries witnessed a series of transformative discoveries that laid the foundation for modern mass spectrometry, including Röntgen's observation of X-rays in 1895 (Röntgen, 1895), Thomson's identification of the electron in 1897 (Thomson, 1897), and Wien's discovery of positive rays, later recognized as protons (Rutherford, 1919). These developments enabled Thomson to construct the first mass spectrometer in 1912 (Thomson, 1912;Thomson, 1913), an instrument subsequently refined by Aston and others (Griffiths, 2008;Münzenberg, 2013). Over the following decades, rapid technological innovation led to the development of time-of-flight analyzers (Wiley and McLaren, 1955), quadrupole and iontrap instruments (Paul and Steinwedel, 1953;Futrell and Miller, 1966;Yost and Enke, 1978), and Fourier transform ion cyclotron resonance mass spectrometry (Comisarow and Marshall, 1974), among many other advancements that define today's high-performance platforms. In recent years, the demand for mass spectrometry instrumentation has grown substantially, driven by expanding applications in clinical diagnostics, pharmaceutical development, biotechnology, environmental monitoring, food safety, and multi-omics research (Seger and Salzmann, 2020;Birhanu, 2023). In particular, the increasing adoption of proteomics, metabolomics, lipidomics, and single-cell analysis has created a need for instruments with higher sensitivity, throughput, robustness, and quantitative accuracy (Momenzadeh and Meyer, 2025). At the same time, major progress has been made in the automation of mass spectrometry workflows, including automated sample preparation, integrated liquid chromatography-mass spectrometry platforms, real-time data processing, and standardized analytical protocols that have accelerated mass spectrometric applications in routine clinical laboratories (Vogeser and Habler, 2026). At the same time, instrument miniaturization has also produced more compact mass spectrometers compatible with field applications such as environmental surveillance, forensic investigations, food authentication, and point-of-care testing (Smith et al., 2024;Pond et al., 2026). Combined, these developments are reshaping the role of mass spectrometry from a specialized research technology into a broadly accessible analytical platform with applications spanning laboratories, clinics, and on-site analyses. To explore these developments and the challenges accompanying them, we brought together representatives from major mass spectrometry manufacturers for a round-table discussion focused on emerging technologies, data standardization, automation, accessibility, and future directions for the field. The discussion also addressed the importance of collaboration among academia, industry, clinicians, and regulatory agencies, as well as education and training for the next generation of scientists. This paper is divided into two parts: Part I (the current paper) covers research applications and technical gaps, including single-cell analysis, metabolomics, lipidomics, and multi-omics integration. And Part II addresses data management and bioinformatics, and collaboration and industry needs. This paper is divided into two parts: Part I (the current paper) covers research applications and technical gaps, including single-cell analysis, metabolomics, lipidomics, and multi-omics integration. And Part II addresses data management and bioinformatics, and collaboration and industry needs. The discussion was published as a webinar, and here we present a summary of that digital event. The discussion was published as a webinar, and here we present a summary of that digital event. The discussion was published as a webinar, and here we present a summary of that digital event. Polarity switching is routinely used in lipidomics, alternating between negative and positive ionization modes during acquisition. Running each mode separately may give us more detailed data, however, we can analyze a single cell only once, otherwise we have to split it in two with one experiment for positive, one experiment for negative ionization, which would mean that we are analyzing half a cell in each mode. Fast polarity switching on modern mass spectrometry instruments helps overcome this by allowing both modes to be captured in a single shot, which significantly improves single-cell lipidomics performance. In bulk samples, where material is abundant, separate runs in positive and negative mode, and improving peak detection by using reverse phase, and HILIC, ion chromatography, gas chromatography are popular, but when we have limited sample amounts, it may not be possible or would decrease performance if we only get one shot. Polarity switching can also help maintain instrument performance. Extended analysis in a single ionization mode can lead to gradual, accumulation of analyte deposits along the ion optics rail, creating a layer that can impede ion transmission into downstream ion optics. Switching polarity helps discharge this buildup and restores sensitivity. Whether we are using fast polarity switching or include it in the method as a washing out step, it really helps the system with fundamental reasons behind it.If a single cell really provides only one analytical shot, integrating multiple omics becomes challenging. The emphasis may be on metabolomics and lipidomics, but proteomics is also important. The broader goal is to extract metabolomic, lipidomic, and proteomic information from the same cell, yet a practical question remains why not using more cells instead of only overlaying multiple pieces of information, such as staining with proteomics or staining with some omics measurement on mass spectrometry. And this is an extremely promising field that will advance our understanding of biology.In some instances, genomics, pathology, proteomics, and small-molecule analysis overlap with multi-omics. The challenge is how to bring these disciplines together into a single data stream capable of describing individuals or populations. Multi-omics tracks the central dogma of biology how genetics affects disease, proteins, small molecule expression, etc. The biggest challenge is merging the data from different fields of science who speak different languages essentially and often have no expertise amongst each other. As analytical chemists, we may understand molecules and biology, but many of us may lack training in bioinformatics or genomic data interpretation, and this gap complicates integration. The more we learn about multi-omics, the clearer it becomes how complex the biological question is.DNA turns into mRNA, which is translated into proteins, which get modified post translationally, degraded and regulated in spatial contexts. Metabolites are generated through networks of pathways rather than simple one-enzyme-one-product synthesis. As we move from genomics to transcriptomics, proteomics, and metabolomics, we can witness a massive expansion of information, and without shared identifiers and standardized ways of describing these data, integration becomes nearly impossible. Maintaining the solid motivation to connect these domains, many companies now develop workflows spanning multiple omics layers. Emerging AI and deep-learning approaches offer promise for making sense of these massive, complex datasets, enabling new insights through advanced visualization and pattern recognition. AI may certainly help in unbiased and unsupervised data analysis. For example, in untargeted analysis, PCA plots are unsupervised, however, some people also prefer a PLS-DA, which is supervised. There are various advantages and disadvantages of those two approaches, and one of the promises of AI is that it can create new models or new understandings of what dependent and independent variables might be by cross correlations across these seemingly unrelated data, which certainly opens new opportunities. AI could be a powerful tool for normalizing data sets so they can be correlated. Proteomics measures milligrams of protein, lipidomics uses millilitres of plasma or another matrix and extract the compound. AI will probably help deal with the massive data sets approaching terabytes in a single sample, but there should be a strategy to assist with fundamental issues, such as normalizing the data sets, so that they can be correlated, then having a unified way of comparing proteins to lipids to which AI may offer some exciting solutions. One point that emerged in the discussion is that transcriptomics and proteomics data sometimes do not correlate perfectly. The familiar model of one gene producing one mRNA that produces one protein may be simplistic when evaluated from the multi-omics perspective. In practice, even when a protein is known to catalyze the formation of specific lipid metabolites, the expected downstream lipid changes may not appear, which may be because numerous additional genes and proteins interplay in the pathways. In such networks, AI could assist by identifying the relevant pathway among the possibilities. Hence, integrating AI tools into mass spectrometry workflows could be a useful development.An example of multilateral biochemical pathways could be the studies in eicosanoids, where phospholipids release arachidonic acid, which cyclooxygenase converts into prostaglandins, a mechanism that underlies the action of NSAIDs, such as Advil and Tylenol. However, in certain tissues, an alternative route exists where arachidonic acid is first converted into 2arachidonoylglycerol and later hydrolysed to eicosanoids. This showed that biochemical pathways may interconnect through various biochemical reactions, and their activity may be state and tissue dependent. With advanced supervised or unsupervised learning applied to large datasets, AI's function will likely be a scientific copilot, supporting, not replacing, the researcher. AI still lacks the conceptual understanding of biochemical context and dependency, so human guidance remains essential. But as an augmenting tool, especially in pathway mapping, it could be transformative.The analytical approach we choose fundamentally shapes our outcomes. In small-molecule analysis, early biochemical pathways were elucidated without mass spectrometers (relying mainly on chemists), but today, targeted analyses using triple-quadrupole mass spectrometers are foundational. Although broad, discovery-style experiments are powerful, hypothesisdriven targeted measurements are often the most effective way to answer specific biological questions. After years of trials, we may have realized that asking the question what was in the sample may not be the correct query. Instead, starting with a targeted sequence, while asking the questions whether compound A, B, or C is present may ultimately lead to the correct answer. Lipidomics illustrates this shift well. Early shotgun approaches were wide and semitargeted, but the field has increasingly moved toward targeted approaches because current software still struggles with isomers, isobars, and complex data processing. Proteomics is further ahead, but metabolomics and lipidomics remain challenging as we continue refining how to measure these molecules accurately.AS: Conceptualization, investigation, writing original draft, data curation, supervision. PB:Writing, reviewing and editing, data curation. PRSB: Writing, reviewing, editing, data curation. LBM: Writing, reviewing, editing, data curation.The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.
No takes yet. Share an insight, caveat, or question.
Szeitz et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: