Pain induces a rapid characteristic metabolic signature detectable in breath

Pain induces a rapid characteristic metabolic signature detectable in breath | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Pain induces a rapid characteristic metabolic signature detectable in breath Mélina Richard, Kapil Singh, Dilan Sezer, Sarah Buergler, Luana Palermo, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6048423/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 21 Apr, 2026 Read the published version in iScience → Version 1 posted You are reading this latest preprint version Abstract The objectification of pain presents a significant clinical challenge, particularly in children, elderly individuals, patients with disabilities and unconscious patients. It is critically important to accurately assess pain in these populations due to the heightened risk of undertreatment. Using the cold pressor test (CPT) as a pain induction model, we combined real-time breath metabolomics with pathway analysis to uncover metabolic shifts. Exhaled breath was analyzed in a discovery cohort (n=19) and validated in an independent cohort (n=21) using secondary electrospray ionization-high-resolution mass spectrometry (SESI-HRMS). Within 15 minutes of CPT, over 400 conserved mass spectral features were significantly altered across both cohorts. Pathway analysis highlighted shifts in aminoacyl-tRNA biosynthesis, cysteine/methionine metabolism, butanoate metabolism, and arginine/proline metabolism. Arginine and glutamate, key contributors to nitric oxide production and nociceptive signaling, exhibited consistent upregulation. Neural network classifiers achieved robust differentiation between pre- and post-CPT profiles (AUC=0.856), showcasing breath metabolomics as a promising observer independent, and objective tool for real-time pain assessment. To validate universal mechanistic relevance of the findings, we compared them to findings of chronic pain studies revealing consistencies in amino acid and neurotransmitter-related pathways. This study provides novel insights into the metabolic basis of acute pain and positions breath metabolomics as a viable approach for dynamic, observer independent monitoring pain in vulnerable patient groups. Future research must determine if these new insights into mechanistic pathways can inform patient- and disease-specific pain management strategies. Health sciences/Biomarkers/Diagnostic markers Health sciences/Medical research/Translational research Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Pain, in its broad range of severities, remains to be a major human and healthcare systems’ burden 1 , 2 . Pain is the most frequent cause of admissions to emergency facilities and is also commonly experienced in primary healthcare settings 3 – 5 and significantly contributes to health care costs. Straightforward strategies for lowering costs in healthcare is acute pain management, and several nations have published recommendations or evidence syntheses on pain management 6 , 7 . Nevertheless, pain assessment is often based on observer dependent measures of pain and related scoring systems. Despite these efforts, the objectification of pain presents a significant clinical challenge particularly in vulnerable risk groups. This is particularly true in children 8 , elderly individuals 9 , patients with cognitive impairment 10 and unconscious patients 11 . It is critically important to accurately assess pain in these populations due to the heightened risk of undertreatment. E.g. pain assessment in children depends on factors such as cognitive development, clinical context, and pain typology, with self-reporting used for children over 6 years and behavioral scales for younger children 8 . As an example of objective pain assessment, studies have explored skin conductance (SC) as a potential tool for measuring acute pain. While some findings suggest SC may be a useful indicator, others have reported inconsistent results 12 – 14 . In patients with cognitive impairment, the undertreatment of pain remains a significant issue in healthcare. 10 In elderly patients, healthcare professionals tend to underestimate pain needs and under-prescribe medications due to lack of objective measures 9 . Even more challenging is the pain assessment and management in unconscious ICU patients due to communication barriers. Unrecognized or undertreated pain affects 70% of ICU patients, leading to serious physical and psychological consequences 11 . Furthermore, in the operating theatre, there is a need for rapid assessment of changes in pain during a surgical intervention or in the recovery room 15 , 16 . In the context of personalized health, we lack the necessary knowledge to tailor evidence-based approaches for individual patients or even specific subclasses. Most human research is grounded in population-based outcomes, yet the reality is that many patients respond to interventions in ways that differ significantly from the average 17 . Hence it is important to investigate further into underlying molecular mechanisms associated with pain responses. We hypothesize that characterizing the exhaled metabolome is ideally suited for these four challenges such as the need for an objective observer independent pain metrics , the suitability for vulnerable risk groups , the possibility to detect immediate changes in pain during interventions and the assessment of metabolic response profiles enabling potentially personalized treatment regimes in the future). Exhalomics offers the possibility to detect rapid —in the time scales of minutes 18 —fluctuations in concentrations as a result of a stressor such as acute pain. Metabolites have also been demonstrated to have exceptional predictive abilities and to closely mirror the actual phenotype 19 , 20 , providing an important path towards personalized medicine 21 , 22 , including pain 23 , 24 . Among the whole metabolome, the most volatile subset of it is thought to play a key pain-signaling role in the animal kingdom: olfactory cues may convey pain to others 25 . In this proof-of-concept human experimental study in two cohorts, we propose a real-time, non-invasive assessment of biochemical changes of experimentally induced pain by cold pressor test (CPT) by harnessing the metabolome via exhaled breath. The CPT is the most popular pain-provocation test in history 23 , 26 . It entails keeping one hand submerged in ice water until it becomes intolerable, at which time the hand is removed 26 . The objective of this study was to assess whether there is a CPT-induced metabolic signature that may allow us to gain further insights into the molecular mechanisms of pain. To this end, we applied a well-established, non-invasive and real-time breath metabolomics pipeline first in a discovery cohort, and subsequently in a validation cohort. Results Study design A total of n = 20 healthy participants were enrolled in the University Children’s Hospital Basel in Switzerland ( i.e. , discovery cohort). One participant withdrew the consent retrospectively, hence n = 19 participants’ data were analyzed. A validation cohort of n = 21 healthy participants were recruited within Jinan University in China ( i.e. , validation cohort; Fig. 1 ). The demographic characteristics of the discovery and validation cohorts were compared. The median age of participants in the discovery cohort was 26.0 years (interquartile range [IQR]: 5.5 years), while the median age in the validation cohort was 24.0 years (IQR: 5.0 years). A Mann-Whitney U test revealed no significant difference in age distribution between the two cohorts (U = 187.0, p = 0.744). Regarding gender distribution, the discovery cohort included 7 males and 12 females, while the validation cohort comprised 12 males and 9 females. A chi-square test found no significant difference in gender distribution between the cohorts (χ² = 0.935, p = 0.334). The N = 40 participants of both sites completed four exhalation measurements, two before- (-15 and − 5 min) and two after-CPT (+ 0 and + 25 min) intervention. A multifaceted univariate and multivariate data analysis pipeline was deployed to identify altered metabolites/metabolic pathways because of induced pain, as well as to predict whether the breath mass spectral fingerprint corresponds to a pre- or post-CPT sample. Pain-induced physiological and metabolic alterations The pain threshold, as assessed by the time the participants could withstand their hand in cold water, varied substantially across individuals, as reported in literature 27 , 28 . The median (interquartile range) withstanding the hand in iced water was 66 (200) and 36 (24) s for the discovery and validation cohort, respectively. Such intervention resulted in a significant (p < 0.001) increase in blood pressure ( Supplementary Fig. S1 ) . A total of 5,058 mass spectral features were detected in both cohorts, of which 1,377 could be mapped to at least one metabolite from the Human Metabolome Database 29 ( Supplementary Table 1 ). Whether pain induction came accompanied by a sizeable change in the overall exhaled metabolic profile was initially assessed by partial least squared-discriminant analysis (PLS-DA). A 10-fold cross-validated analysis revealed a robust classification performance for both cohorts, as assessed by the accuracy (discovery: 0.97; validation: 0.93) and discriminant Q 2 30 (discovery: 0.74; validation: 0.84). Figure 2 a shows the resulting score plot based on SESI-HRMS breath mass spectra from the participants pre- and post-CPT. A clear shift of the data points upon the intervention is observed in both the discovery and validation cohorts, suggesting that CPT induces a consistent metabolic shift. To gain further insights into the strength and significance of such changes at the molecular level, we conducted a volcano plot analysis (Fig. 2 b). The figure indeed displays a striking change in the metabolic profile after pain induction, whereby the signal intensity of most of the features was increased upon intervention, as evidenced by the asymmetric volcano plot. We compared the area under the curve of both measurements before pain against the area under the curve of both measurements after pain (i.e. measurements 1 + 2 versus 3 + 4). A total of 1,153 features were found to be significantly upregulated (Log 2 FC ≥ 1.5 & FDR ≤ 0.01) in the discovery cohort. In contrast, just 18 features were significantly decreased (Log 2 FC ≤ -1.5 & FDR ≤ 0.01) after pain induction. The analysis of the validation cohort revealed a remarkably similar picture, which exhibited 790 upregulated features and 13 downregulated features, respectively. Interestingly, 416 upregulated and 6 downregulated features overlapped between the two study sites. The mass spectra for all participants of one such example, which was mapped to arginine, is displayed in Fig. 2 c. It shows a generalized and significant increase as a response to the CPT across all subjects for the discovery (mean Log 2 FC = 1.90; FDR < 0.001 ) and validation (mean Log 2 FC = 1.58; FDR < 0.001) cohorts (Fig. 2 d ) . Supplementary Fig. S2 shows additional examples exhibiting a similar trend for creatine, threonine and asparagine. To gain further insights into whether the relationships across the metabolic features were conserved in the discovery and validation cohorts, we conducted a correlation analysis. Figure 3 a shows the resulting correlation network (ρ cut-off > 0.7 in both sites). The color coding of the nodes signifies the mean fold-change (FC) on Log 2 scale (i.e. Log 2 FC).. It becomes apparent that the features experiencing the strongest response after pain induction tend to cluster together and are in the periphery of the network, whereas those with the lowest FC tend to also correlate and are at the core of the network. A visual inspection of the color distribution across both networks provides a sense of resemblance, indicating that also at that level, the data structure is conserved in the validation set. Further analysis of discovery and validation networks across various ρ cutoffs reveals significant structural similarity between the two datasets (Fig. 3 b). The correlation matrix distance 31 of 0.678 (p-value < 0.001) indicates that the networks are highly similar. An eigenvalue correlation of 0.992 (p-value < 0.001) confirms that the overall variance structures of the two networks are almost identical. However, the moderate correlation of correlations (0.187, p-value < 0.001) and the Jaccard Index, which decreases from 0.59 at ρ = 0.1 to 0.06 at ρ = 0.90, suggest that differences in edge patterns and connectivity become more pronounced as weaker correlations are pruned. The weighted edge correlation, ranging from 0.22 to 0.31, reflects moderate consistency in edge weights, further underscoring a partial overlap in network connectivity. The edge difference metric, which decreases steadily from 0.305 at ρ = 0.1 to 0.023 at ρ = 0.9, highlights the reduction in the magnitude of edge weight differences between the two networks as weaker correlations are filtered out. This indicates a progressive alignment in the remaining stronger connections. Degree-FC correlation is consistently negative, becoming more pronounced at higher ρ cutoffs, with values ranging from − 0.074 (p-value < 0.001) to -0.113 (p-value < 0.001) in the discovery cohort and from − 0.220 (p-value < 0.001) to -0.250 (p-value < 0.001) in the validation cohort. This suggests that higher-degree nodes tend to exhibit smaller fold changes, reinforcing the role of hub nodes in maintaining network stability despite local differences. Both networks remain fully connected up to a ρ cutoff of 0.6, beyond which the number of connected components increases rapidly, indicating network fragmentation. The diameter of the networks remains stable at lower ρ values, with a value of 2, but begins to increase as the networks become sparser, reflecting reduced global connectivity. Metrics such as the clustering coefficient, which decreases from 0.38 to 0.05, and the average path length, which increases from 1.27 to 5.56, remain stable at lower ρ cutoffs but diverge as ρ increases. This divergence indicates that local connectivity and efficiency diminish as weaker edges are pruned, further contributing to network sparsity. Betweenness centrality shows a moderate decrease, from 679 to 480, as ρ increases, suggesting a shift in the importance of key nodes in maintaining overall connectivity. This reflects changes in the centrality landscape of the networks as they become fragmented. Overall, while local connectivity and specific edge patterns vary significantly with increasing ρ cutoffs, the global structural similarity between the discovery and validation networks suggests that the networks are underpinned by similar underlying metabolic processes. Pain-induced altered metabolic pathways The univariate, multivariate and correlation network data analysis approaches presented above suggest a stark and rapid metabolic perturbation as a result of the CPT intervention consistent in both cohorts. To generate further hypothesis into the specific altered regions of human metabolome, we conducted a pathway-based degree analysis. A total of 55 metabolic pathways were associated with the 5,058 mass spectral features overall detected in both cohorts ( Supplementary Table 1 ). To increase the robustness of the analysis, we further considered only those pathways represented by at least five mass spectral peaks, reducing the list to 32 pathways. Figure 4 a and Fig. 4 b show an excellent overall agreement for the median Log 2 FC and degree between the discovery and validation cohorts. Further insights into the intra-pathway correlation for the Log 2 FC and the degree could be gained by conducting a hierarchical cluster analysis Fig. 4 c. It emerges from this analysis that the most consistent behavior between the discovery and validation cohorts for both, the Log 2 FC and the degree, was found for alpha-linolenic acid metabolism (ALA), fructose and mannose metabolism (FMM), ketone body metabolism (KBM), butyrate metabolism (BUT), arginine and proline metabolism (APM), aspartate metabolism (ASP), cysteine metabolism (CYS), and histidine metabolism (HIS). These pathways collectively highlight the body’s ability to convert and utilize fatty acids, carbohydrates, and amino acids for energy homeostasis and biosynthetic needs 32 , 33 . To complete the biological interpretation of CPT-driven metabolic shifts, we conducted an enrichment analysis using the mummichog algorithm 34 . We considered as significant feature list those that were significantly altered in both study sites (i.e., 8% of the total feature list; 416/6 upregulated/downregulated; Supplementary Table S1 ). The top five pathways (p gamma < 0.1) include Aminoacyl-tRNA biosynthesis, Cysteine and methionine metabolism, Butanoate metabolism, Alanine, aspartate and glutamate metabolism and Arginine and proline metabolism ( Supplementary Fig. S3 ), comprising a total of 58 compounds mapped to these pathways ( Supplementary Table S2 ). Figure 4 d shows a schematic representation of the how the main identified pathways are interconnected to provide an overview of the main metabolic alterations driven by CPT. Neural networks predict whether a breath mass spectral fingerprint is associated with pain The data presented above suggests rapid and stark metabolic changes taking place as a result of CPT intervention. Therefore, we attempted to establish a neural network model to test the hypothesis of whether a first-line breath test could determine whether a patient is in pain. The resulting ROC curve (Fig. 5 a) summarizes the prediction performance whereby the AUC = 0.856 and the overall model accuracy was 78%. When analyzing the results at an individual level (Fig. 5 b), we observe that the majority of the participants were correctly classified whereby the first two measurements ( i.e. , pre-CPT) were assigned to belong to the “pain” mass spectral signature with a low probability and the last two measurements ( i.e. , post-CPT) with a high probability. However, some individuals were systematically misclassified whereby the second measurement was misclassified as “pain” and the third measurement as “no-pain”. The hierarchical clustering analysis revealed three main clusters. The first cluster predominantly consisted of samples classified as "no-pain," including most pre-CPT measurements, suggesting that their metabolic response to pain induction was either delayed or less pronounced. The second cluster comprised post-CPT samples that were consistently classified as "pain," reflecting a robust metabolic shift induced by the CPT intervention. The third cluster contained a mix of pre- and post-CPT samples, where some pre-CPT samples were misclassified as "pain." This misclassification may point to individual variability in baseline metabolic states, potentially influenced by pre-procedural nervousness or stress, as the CPT intervention might have been anticipated. Overall, the correct classification rate was 92.5%, 70%, 70% and 87.5% for the first, second, third and fourth breath measurement, respectively. Such misclassifications were observed both in the discovery and the validation cohorts. Discussion In this proof-of-concept study, we presented and validated a novel observer independent method which can reliably detect pain in human adults with an accuracy of 78% and an area under the curve of 0.86. The study was not designed to directly compare this novel method to standard clinical pain scores such as Numeric Rating Scale (NRS), Visual Analog Scale (VAS) or Verbal Rating/Descriptor Scale (VRS/VDS). It moreover offers and validates a novel technology which must be tested in future studies in various patient groups under various clinical conditions. Nevertheless, the novel observer independent technology offers the possibility to be applicable to vulnerable risk groups of patients such as infants, children, the elderly, patients with cognitive impairment or sedated or unconscious patients in ICU and operating theatre. In the latter setting, the possibility to detect immediate changes in pain during interventions makes the techniques potentially extremely useful to guide intervention and analgetic regimen during anesthesia. Furthermore, the assessment of metabolic response profiles enabling potentially personalized treatment. Pain is a highly complex sensation which is objectively difficult to describe and likely has a high individual characteristic dependent on disease and circumstances. Research on skin conductance (SC) as a tool for assessing acute pain in infants and children has yielded mixed results. While some studies found SC to be a potentially useful indicator of pain, others reported inconsistent findings. Hullett et al. demonstrated that SC could accurately predict the absence of moderate to severe pain in postoperative pediatric patients 12 . Similarly, Dalal et al. found that peak SC values may serve as indicators of unmitigated pain in infants 13 . However, Solana et al. 35 concluded that SC was not more sensitive or faster than clinical scales for pain assessment in critically ill children. Hu et al. reported inconsistent validity evidence for SC in pain assessment, noting significant correlations with unidimensional behavioral pain scales but not with multidimensional measurements 14 . Overall, while SC shows promise as a pain assessment tool, further research is needed to establish its diagnostic accuracy and clinical applicability across different pediatric populations and pain contexts. The International Association for the Study of Pain has defined pain as “an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage” 36 . Characterizing such a sensory and emotional experience at a molecular level is an inherently challenging task. This pain response involves the activation of the sympathetic nervous system (SNS), which increases sympathetic tone across various organs. The intensity of this pain-related stress response is proportional to the severity of the injury and the level of pain experienced. The body's reaction to pain serves to mobilize energy resources and prioritize vital functions needed for survival. We showed that the bodies response to pain results in changes in metabolism expressed in the changes in exhaled metabolic profiles. These profiles may inherit properties of a patient specific pain response and could potentially help to guide future treatment strategies. These exhaled metabolic profiles were assessed in a non-invasive manner as follows: We used CPT as a proxy for pain induction combined with a well-established real-time breath metabolomics pipeline, applying it in both a European (discovery) and a Chinese (validation) cohort. Our results suggest that pain, as induced by CPT, triggers within 15 min a significant upregulation of ~ 400 metabolic mass spectral features, highly conserved across these populations. Subsequent enrichment and pathway-degree analysis flagged coordinated metabolic adjustments in aminoacyl-tRNA biosynthesis, cysteine/methionine metabolism, butanoate metabolism, alanine/aspartate/glutamate metabolism, and arginine/proline metabolism. Enhanced aminoacyl-tRNA biosynthesis likely supports stress-induced protein synthesis demands, as aminoacyl-tRNA synthetases and related factors can be upregulated under cellular stress 37 . Meanwhile, the upregulation of cysteine and methionine metabolism can reflect both increased antioxidant defenses and shifts in methylation crucial for redox control and possibly catecholamine turnover 38 . Butanoate metabolism may help supply alternative energy substrates and modulate inflammation during acute stress, supported by evidence that short-chain fatty acids affect host immune and metabolic homeostasis 39 . In parallel, alterations in alanine, aspartate, and glutamate metabolism—particularly glutamate’s central role in nociceptive transmission—could reinforce pain signaling in the central nervous system 40 . Finally, the arginine–proline axis, which governs nitric oxide (NO) production 41 and tissue remodeling, may synergize with sympathetic activation to fine-tune vascular tone, culminating in the hallmark blood pressure increase during CPT. This high-level metabolic shift aligns with prior work showing CPT-induced alterations at both transcriptome and metabolome levels, reflecting SNS activation in stress and pain responses 23 , 42 . For example, arginine was previously found to correlate with gene expression changes following CPT, consistent with our observation of elevated arginine (average Log 2 (FC) = 1.90 in the discovery cohort; 1.58 in the validation cohort; Fig. 2 c and Fig. 2 d), a key amino acid in NO production and vascular regulation. Additionally, arginine plays a role in energy mobilization, supporting the body's efforts to prioritize survival functions when the SNS is activated, especially during stress and pain 43 , 44 . Glutamic acid also increased (average Log 2 (FC) = 1.74 in discovery; 1.44 in validation) and, as an excitatory neurotransmitter, may intensify nociceptive signaling while impacting blood pressure regulation 45 – 47 . Gamma-aminobutyric acid (GABA), in contrast, generally acts inhibitory but showed elevated levels as well (average Log 2 (FC) = 1.53 in discovery; 1.37 in validation). Dysregulation in GABA balance can contribute to heightened SNS excitability, pain perception, and blood pressure. Creatine (average Log 2 (FC) = 1.62 in discovery; 1.92 in validation; Supplementary Fig. S2 ) further supports energy metabolism during heightened SNS activation 48 . Interestingly, these findings parallel meta-analyses on chronic pain that report alterations in amino acids (e.g., glutamine, serine, phenylalanine) and macromolecule metabolism intermediates (e.g., succinate, citrate, acetylcarnitine, and N-acetylornithine) 49 . When we look further into cold induced pain, a recent meta-analysis on metabolomics in chronic pain research 50 has found, that pathways resulting in increased glutamate levels, either systemically or locally, have been exhibited in a variety of conditions, including fibromyalgia, Complex regional pain syndrome, trapezius myalgia, and mixed musculoskeletal pain conditions. It is widely recognized that glutamate signaling plays a part in both central and peripheral sensitization 51 , 52 , 53 . Although the type of pain in our study was not chronic, this constancy holds true for many types of pain. Furthermore, the sensitivity to decreasing temperature is significantly increased among chronic pain patients throughout conditions 54 . Glutamate metabolites may serve as indicators of nociception in a variety of pain disorders that also show in cold receptors potential where cold temperatures and irritating substances can stimulate the transient receptor potential TRPA1 channel in sensory neurons 55 , 56 . Furthermore, TRPA1 stimulation enhanced glutamate release onto the nucleus tractus solitarii in the literature 57 , which points once more to the direction of the involvement of glutamate related mechanisms. Altogether, our data underscores the importance of amino acid and carbohydrate metabolism for immediate energy needs, alongside lipid and redox pathways for sustained adaptation. This dynamic prioritization of metabolic pathways reflects the body's adaptability to acute stress 58 , 59 . Beyond the importance of understanding such molecular processes triggered to cope with pain and stress, our neural network predictions (Fig. 5 ) suggest that such rapid shifts in exhaled metabolic patterns could potentially be harnessed in a clinical context to gauge pain more objectively than current practice. It is important to acknowledge and consider several limitations associated with this study. Firstly, the inclusion of a relatively small number of sites and subjects compared to the generated abundance of datapoints in -omics fields, may have led to underpowered statistical analyses. Secondly, the identification of metabolites in breath samples relied on database matching using accurate mass measurements within a narrow range of ± 1 ppm. While this approach provides a preliminary identification of metabolites, it lacks the comprehensive characterization that can be achieved through mass separation techniques. Consequently, caution should be exercised when interpreting the biochemical significance of the identified metabolites. Further validation using additional analytical methods, such as tandem mass spectrometry or nuclear magnetic resonance spectroscopy, would enhance the confidence and accuracy of the metabolite identifications. Thirdly, pain manifests in various forms, and without additional studies, it remains uncertain whether our findings are exclusively attributable to pain. Other factors, such as cold exposure, stress, fluctuations in blood pressure, and related physiological responses, may also contribute. However, this overlap also serves as a means of defining pain within a broader physiological and contextual framework. Fourthly, the specific involvement of metabolic pathways and metabolites in receptor activators and pain processing is grounded in sparse and possibly underpowered studies and their precise roles in pain modulation, nociception, and noxious cold sensation require further investigation. Fifthly, the study and understanding of metabolic pathways within pain-related contexts are still in their nascent stage. Currently, numerous potential metabolic feature hits pertaining to this subject matter are being published. However, it is yet to be determined which specific findings will emerge as definitive and significant contributions, as further investigations and scientific advancements are required to unravel the complexities of these pathways in the context of pain, and in detangling the complex nature of pain itself. Having said this, one main strength of our study in this regard lies in the fact that we included a validation cohort in an independent setting, which revealed a remarkably similar response to pain induction via CPT. The highly standardized procedures conducted in both laboratories ensured a minimization of the technical variability along with a high mass accuracy 60 , 61 . Moreover, several central metabolites relevant to this study (e.g. amino acids 62 and citric acid cycle 63 ), have been previously fully characterized in exhaled breath. In summary, we show here that CPT-driven pain induces a rapid physiological response encompassing a stark metabolic shift within time scales of minutes. This was observed leveraging the potential of real-time high resolution mass spectrometry to interrogate the exhaled metabolome. The altered metabolome illustrates the complex interplay between the sympathetic nervous system, pain response, energy mobilization and blood pressure regulation. Through their various roles in neurotransmission, vascular control, and energy metabolism, they help the body adapt to stressful or painful situations. Taken together, these interlinked pathways appear to optimize energy utilization, augment neurotransmission, and balance vasoactive mediators, thus promoting both the acute pain response and the hemodynamic elevation characteristic of CPT. Finally, we confirmed that such changes in the exhaled metabolic fingerprints could be used to provide non-invasively and in real time a probability score of whether the subject was in pain. These observations were observed intially in a European cohort and further validated in a Chinese cohort, reinforcing the robustness of our observations. We hypothesize that further development of this observer-independent technique could potentially guide personalized pain management in a clinical context in the various vulnerable risk groups of patient in a chronic or acute pain setting. This method could help to avoid undertreatment in these groups of patients. Methods Clinical Trial Design: Pain induction via CPT Pain was experimentally induced by the CPT 64 – 66 . During the trial, participants underwent four exhalation measurements (Fig. 1 ): Two baseline sets before the CPT, one immediately after the hand withdrawal from the cold water, and a subsequent final washout measurement 15 minutes after the third measurement. Participants submerged their open and relaxed right hand and 3 cm of their wrist into a container of around 13 liters containing ice-cold water and crushed ice that was manually stirred and continuously measured for temperature of 2°C (± 1°C) using a laboratory thermometer. Participants were instructed to keep their hand and wrist submerged in the water for as long as they could tolerate the pain before removing their hand, however, a 4-minute cutoff was applied, and those who could manage to keep their hand submerged for that long were also instructed to remove it. The participants received the study information and consents by email at least 24 hours before the experimental intervention. The two information packages provided to participants described the trial in simpler, shorter, and more complex detailed versions. At the study site, the entire procedure was explained in detail, and participants’ questions were addressed before obtaining consent. Informed consent and general consent were obtained from the participants before starting measurements. Before the start of the project, ethical approvals (Swiss equivalent to institutional review boards IRBs) were registered and approved with the following endpoints. Primary endpoint: The variable of primary interest is the variation of signal intensity of mass-to-charge ratio ( m/z ) of exhaled metabolites across the conducted CPT pain analgesia-placebo effects trial. Secondary endpoints: Differentially exhaled metabolites across the pain vs. no-pain measurements. Exhaled metabolites for differences between placebo responders and non-responders. Exhaled metabolites association with blood pressure and heart rate. Pre-registrations and ethical approvals are listed in Supplementary Table 3 . Real-time breath metabolomics by High-Resolution Mass Spectrometry The breath metabolomics platform consisted of a Secondary Electrospray Ionization-High Resolution Mass Spectrometry (SESI-HRMS) setup. For both study sites the same settings were chosen if not mentioned separately: The system involved a direct coupling of a high-resolution mass spectrometer (Q-Exactive Plus in the Swiss site and Q-Exactive in the Chinese site, Thermo Fisher Scientific, Germany) to an ion source (Super SESI, Fossiliontech FIT, Spain). A 20 µm capillary emitter (Fossiliontech FIT, Spain) was employed to generate an electrospray of water (0.1% v/v formic acid; LiChrosolv, hypergrade for LC-MS, 1.59013.2500, Millipore). The pressure within the electrospray reservoir was maintained at a pressure of 1.3 bar for the Swiss site and 0.8bar for the Chinese site respectively. Electrospray voltages were set at 2.8 kV in positive and negative mode. The ionization chamber and sample line were set to temperatures of 90°C and 130°C, respectively. The Orbitrap capillary temperature was set to 275°C, the sheath gas flow rate was set at 60 AU, and the S-lens RF level was set to 55.0 V. To capture mass spectra, the Thermo Exactive Plus Tune software (5.0.0.38 for the Swiss site and 2.8.1.2806 for the Chinese site) was utilized in full scan mode with a resolving power of 140,000 at m/z 200. The scan parameters included a range of 70-1000 m/z, selection of positive or negative polarity, 2 microscans, an ACG target of 1e6, and a maximum injection duration of 500 ms. The MS was internally calibrated by enabling lock masses corresponding to typical background mass spectrometric contaminants 67 68 , and externally calibrated on a weekly schedule using a commercially available calibration solution (PierceTM Triple Quadrupole, extended mass range) 69 . Prior to any breath measurement, an external gas standard (α-terpinene at 100 ppb | Dalian Special Gases Co. Ltd, Dalian, China) 70 was used to test the instrument's sensitivity. The so-called Nelson criteria in process control were employed to validate if the instrument performance was within the “in-control” range using historical data on the same gas standard 71 . Once this suitability test was passed, the subjects were invited to perform the breath test. The method followed the standardized approach described previously 60 , 72 . In short, an exhalation set includes six exhalations in positive and six exhalations in negative mode. The exhalations were controlled by a mass flow controller and guided by CO 2 exhalation. The process of one set took approximately seven minutes. Data analysis Pre-processing of mass spectrometric data The preprocessing of file spectra was undertaken as per our patented pipeline 73 , 74 : The acquired *.RAW file spectra were preprocessed using the in-house developed SESI-HRMS_Analysis_Toolbox (v 5.2.0) in MATLAB (version R2023a,b-2024a,b; MathWorks Inc., USA) and C# environment. In summary, exhalation time windows including only the end-tidal fraction were determined by the intervals when CO 2 concentrations were greater than 3% (as measured by Exhalion capnograph). Average raw centroid and profile mass spectra during such exhalation scans were computed using in-house C# console apps based on Thermo Fisher Scientific’s RawFileReader (version 5.0.0.38). Centroid and profile mass spectra were recalibrated to achieve mass accuracies within ± 1 ppm. Subsequently, apodization of artifact satellite peaks 75 was accomplished. Finally, the centroid data set was binned within ± 1 ppm using MATLAB’s ksdensity function. This resulted in a final feature list of size 11,840 in positive mode and 4,446 in negative mode. Post-processing of mass spectrometric data Firstly, we reduced the sparsity of the data matrix by dropping all mass spectral features with at least 50% of zero values across all samples. This reduced the number of features to 3,734 and 1,324 in positive and negative mode, respectively (i.e., 5,058 features overall). Secondly, the remaining zeros were imputed using the regression on order statistics method 76 . To identify rapid changes in breath composition as result of CPT intervention, the two baseline and the two post-CPT measurements were fused by calculating the area under the curve (AUC) between the two pairs of measurements and normalized by the time between both measurements. Thereafter, the Log 2 FC between the pre- and post-CPT AUCs was computed for each feature. Log 2 FC were subjected to a paired t-test. Significant features defined as those for which their positive false discovery rate 77 (i.e., FDR) ≤ 0.01 and the |Log 2 FC| ≥ 1.5. Biological interpretation: database query, correlation, and enrichment analysis We matched the final features to molecular formulas with adducts M + H [1+], M(C13) + H [1+], M-H [1-], and M(C13)-H [1-] and a ppm error window of +-1, using a list that contains the “Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry all possible molecular formulas“ 78 . Further, we utilized the Human Metabolome Database (HMDB) data set ‘serum metabolites (5.0)’ 29 to map our molecular formulas onto HMDB accession, taxonomy class, pathways, and metabolite descriptors. Correlation networks were computed using Pearson correlation and visualized using Cytoscape 79 upon thresholding to rho > 0.7. Using this criterion, we proceeded to identify pairs of nodes that satisfied the threshold condition in both study sites. Only those node pairs whose correlations exceeded the cutoff in both matrices were selected for further analysis. The resulting pairs of nodes, fulfilling the aforementioned condition, were then organized into a tabular format representing the nodes in interaction that exhibit strong correlations across both discovery and validation study sites. Pathway enrichment analysis was performed using mummichog algorithm implemented in MetaboAnalystR (version 4.0.0) using hsa_kegg database and abovementioned adduct and tolerances. In cases of overlapping peaks, we assigned a lower probability weight of 0.001 to the corresponding entries in the vector dummy Ps which was otherwise initialized to 1 for all data points, ensuring differentiation between unique and overlapping signals. Pain prediction using neural networks The prediction of pre-CPT vs post-CPT (i.e., no pain vs. pain) breath mass spectra was accomplished using a multi-layer perceptron network in a Python environment. Firstly, the profile mass spectra were linearly interpolated, and the positive and negative ion mode spectra concatenated. This resulted in a final data matrix of 160 samples (2 samples pre-CPT + 2 samples post-CPT = 4 breath samples per subject; 19 participants in discovery cohort and 21 in validation cohort). The dimensionality was reduced via principal component analysis (PCA) to 48, 64, 96 128 or 144 features. The number of input neurons of the network architecture corresponded to the available PCA features followed by two hidden layers where the sizes: [32,32], [48,48], [64,64] or [96,96] where evaluated. The data − at the participant level − was split in 90% training, 5% validation and 5% testing for cross validation. This process was repeated 50 times with random partitions in each fold and for all combinations of number of PCA features and sizes of hidden neurons. Thus, all four breath mass spectra for a given participant would be either in the training, validation or the test set but never mixed to avoid potential classification bias. The training set was subjected to PCA to reduce the dimensionality, whereby the first n scores were used as predicting features. Once the model was trained, the test set was projected onto the same PCA space of the training set and predicted using the first n scores. The pain/no-pain prediction performance was estimated by the area under the curve of the ROC curve. Upon hyperparameter optimization of the network architecture (number of PCA scores and size of the hidden layer), the best setting classification scores were obtained with 48 PCA features with 96,96 hidden layers. Declarations Competing interests PS is co-founder and board member of Deep Breath Intelligence (DBI) AG, a company that provides services in the field of breath analysis. Kapil Dev Singh is partially employed by the same company. UKBB is a shareholder of said company. Funding This work was funded by grants from Fondation Botnar (Switzerland) No. 320030_173168 (to PS) and the Swiss National Science Foundation (SNSF) PCEGP3-181300/ 501100001711-173168 (to PS). XL received funding from Guangdong Major Project of Basic and Applied Basic Research (No. 2023B0303000013), Guangdong Provincial International Science and Technology Cooperation Project (No. 2022A0505050044), and the National Natural Science Foundation of China (No. 22122603). References Brennan, F., Carr, D. B. & Cousins, M. Pain Management: A Fundamental Human Right. Anesthesia & Analgesia 105 , 205-221 (2007). https://doi.org/10.1213/01.ane.0000268145.52345.55 Gaskin, D. J. & Richard, P. The Economic Costs of Pain in the United States. The Journal of Pain 13 , 715-724 (2012). https://doi.org/10.1016/j.jpain.2012.03.009 Cordell, W. H. et al. The high prevalence of pain in emergency medical care. Am J Emerg Med 20 , 165-169 (2002). https://doi.org/10.1053/ajem.2002.32643 Chou R, W. J., Ahmed AY, et al. Vol. Comparative Effectiveness Review, No. 240. (Agency for Healthcare Research and Quality (US), Rockville (MD), 2020). Finley, C. R. et al. What are the most common conditions in primary care? Canadian Family Physician | Le Médecin de famille canadien Systematic review 64 , 832-840 (2018). Chou, R. et al. Management of Postoperative Pain: A Clinical Practice Guideline From the American Pain Society, the American Society of Regional Anesthesia and Pain Medicine, and the American Society of Anesthesiologists' Committee on Regional Anesthesia, Executive Committee, and Administrative Council. The Journal of Pain 17 , 131-157 (2016). https://doi.org/https://doi.org/10.1016/j.jpain.2015.12.008 Schofield, P. et al. Evidence-based clinical practice guidelines on the management of pain in older people - a summary report. Br J Pain 16 , 6-13 (2022). https://doi.org/10.1177/2049463720976155 Beltramini, A., Milojevic, K. & Pateron, D. Pain Assessment in Newborns, Infants, and Children. Pediatr Ann 46 , e387-e395 (2017). https://doi.org/10.3928/19382359-20170921-03 Catananti, C. & Gambassi, G. Pain assessment in the elderly. Surg Oncol 19 , 140-148 (2010). https://doi.org/10.1016/j.suronc.2009.11.010 Cook, A. K., Niven, C. A. & Downs, M. G. Assessing the pain of people with cognitive impairment. Int J Geriatr Psychiatry 14 , 421-425 (1999). https://doi.org/10.1002/(sici)1099-1166(199906)14:63.0.co;2-y Alderson, S. M. & McKechnie, S. R. Unrecognised, undertreated, pain in ICU—Causes, effects, and how to do better. Open Journal of Nursing 03 , 108-113 (2013). https://doi.org/10.4236/ojn.2013.31014 Hullett, B. et al. Monitoring electrical skin conductance: a tool for the assessment of postoperative pain in children? Anesthesiology 111 , 513-517 (2009). https://doi.org/10.1097/ALN.0b013e3181b27c18 Dalal, P. G. et al. Analysis of acute pain scores and skin conductance measurements in infants. Early Human Development 89 , 153-158 (2013). https://doi.org/https://doi.org/10.1016/j.earlhumdev.2012.09.008 Hu, J., Harrold, J., Squires, J. E., Modanloo, S. & Harrison, D. The validity of skin conductance for assessing acute pain in mechanically ventilated infants: A cross‐sectional observational study. European Journal of Pain 25 , 1994-2006 (2021). https://doi.org/10.1002/ejp.1816 Brown, D. N. Pain assessment in the recovery room. J Perioper Pract 18 , 480-489 (2008). https://doi.org/10.1177/175045890801801103 Breivik, H. et al. Assessment of pain. British Journal of Anaesthesia 101 , 17-24 (2008). https://doi.org/10.1093/bja/aen103 Aldington, D. & Eccleston, C. Evidence-Based Pain Management: Building on the Foundations of Cochrane Systematic Reviews. American Journal of Public Health 109 , 46-49 (2019). https://doi.org/10.2105/ajph.2018.304745 Link, H., Fuhrer, T., Gerosa, L., Zamboni, N. & Sauer, U. Real-time metabolome profiling of the metabolic switch between starvation and growth. Nature Methods 12 , 1091-1097 (2015). https://doi.org/10.1038/nmeth.3584 Rohde, P. D., Kristensen, T. N., Sarup, P., Munoz, J. & Malmendal, A. Prediction of complex phenotypes using the Drosophila melanogaster metabolome. Heredity (Edinb) 126 , 717-732 (2021). https://doi.org/10.1038/s41437-021-00404-1 Zhou, S. et al. Systems genetics of the Drosophila metabolome. Genome Res 30 , 392-405 (2020). https://doi.org/10.1101/gr.243030.118 Castelli, F. A. et al. Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to point-of-care tests. Anal Bioanal Chem 414 , 759-789 (2022). https://doi.org/10.1007/s00216-021-03586-z Jacob, M., Lopata, A. L., Dasouki, M. & Abdel Rahman, A. M. Metabolomics toward personalized medicine. Mass Spectrom Rev 38 , 221-238 (2019). https://doi.org/10.1002/mas.21548 Kogelman, L. J. A. et al. Multi-omics to predict changes during cold pressor test. BMC Genomics 23 , 759 (2022). https://doi.org/10.1186/s12864-022-08981-z Hsu, W. H. et al. Metabolomic and proteomic characterization of sng and pain phenotypes in fibromyalgia. Eur J Pain 26 , 445-462 (2022). https://doi.org/10.1002/ejp.1871 Lötsch, J. et al. Smell of pain: intersection of nociception and olfaction. PAIN 157 , 2152-2157 (2016). https://doi.org/10.1097/j.pain.0000000000000599 Modir, J. G. & Wallace, M. S. in Analgesia: Methods and Protocols (ed Arpad Szallasi) 165-168 (Humana Press, 2010). Lamotte, G., Boes, C. J., Low, P. A. & Coon, E. A. The expanding role of the cold pressor test: a brief history. Clinical Autonomic Research 31 , 153-155 (2021). https://doi.org/10.1007/s10286-021-00796-4 Fanninger, S., Plener, P. L., Fischer, M. J. M., Kothgassner, O. D. & Goreis, A. Water temperature during the cold pressor test: A scoping review. Physiology & Behavior 271 , 114354 (2023). https://doi.org/https://doi.org/10.1016/j.physbeh.2023.114354 Wishart, D. S. et al. HMDB 5.0: the Human Metabolome Database for 2022. Nucleic Acids Res 50 , D622-d631 (2022). https://doi.org/10.1093/nar/gkab1062 Westerhuis, J. A., Van Velzen, E. J. J., Hoefsloot, H. C. J. & Smilde, A. K. Discriminant Q2 (DQ2) for improved discrimination in PLSDA models. Metabolomics 4 , 293-296 (2008). https://doi.org/10.1007/s11306-008-0126-2 Herdin, M., Czink, N., Ozcelik, H. & Bonek, E. in 2005 IEEE 61st Vehicular Technology Conference. 136-140 Vol. 131. Berg, J. M., Tymoczko, J. L., Gatto, J. G. J. & Stryer, L. Biochemistry . Eighth edition, fourth printing edn, (Macmillan [etc.], 2015). Kanehisa, M. et al. Data, information, knowledge and principle: back to metabolism in KEGG. Nucleic Acids Res. 42 , D199-205 (2014). https://doi.org/10.1093/nar/gkt1076 Li, S. et al. Predicting network activity from high throughput metabolomics. PLoS Comput Biol 9 , e1003123 (2013). https://doi.org/10.1371/journal.pcbi.1003123 Solana, M. J. et al. Assessment of pain in critically ill children. Is cutaneous conductance a reliable tool? Journal of Critical Care 30 , 481-485 (2015). https://doi.org/https://doi.org/10.1016/j.jcrc.2015.01.008 Raja, S. N. et al. The revised International Association for the Study of Pain definition of pain: concepts, challenges, and compromises. PAIN 161 , 1976-1982 (2020). https://doi.org/10.1097/j.pain.0000000000001939 Park, S. G., Ewalt, K. L. & Kim, S. Functional expansion of aminoacyl-tRNA synthetases and their interacting factors: new perspectives on housekeepers. Trends in Biochemical Sciences 30 , 569-574 (2005). https://doi.org/10.1016/j.tibs.2005.08.004 Espinosa-Diez, C. et al. Antioxidant responses and cellular adjustments to oxidative stress. Redox Biol. 6 , 183-197 (2015). https://doi.org/https://doi.org/10.1016/j.redox.2015.07.008 Dalile, B., Van Oudenhove, L., Vervliet, B. & Verbeke, K. The role of short-chain fatty acids in microbiota-gut-brain communication. Nat. Rev. Gastroenterol. Hepatol. 16 , 461-478 (2019). https://doi.org/10.1038/s41575-019-0157-3 Latremoliere, A. & Woolf, C. J. Central Sensitization: A Generator of Pain Hypersensitivity by Central Neural Plasticity. The Journal of Pain 10 , 895-926 (2009). https://doi.org/10.1016/j.jpain.2009.06.012 Palmer, R. M. J., Ashton, D. S. & Moncada, S. Vascular endothelial cells synthesize nitric oxide from L-arginine. Nature 333 , 664-666 (1988). https://doi.org/10.1038/333664a0 Clarke, H. E., Kluess, H., Adams, S., Kim, D.-H. & Hickner, R. Metabolomics Profiling in Response to a Cold Pressor Test. The FASEB Journal 34 , 1-1 (2020). https://doi.org/https://doi.org/10.1096/fasebj.2020.34.s1.06129 Seravalle, G. & Grassi, G. Sympathetic Nervous System, Hypertension, Obesity and Metabolic Syndrome. High Blood Press. Cardiovasc. Prev. 23 , 175-179 (2016). https://doi.org/10.1007/s40292-016-0137-4 Hirooka, Y. Sympathetic Activation in Hypertension: Importance of the Central Nervous System. Am. J. Hypertens. 33 , 914-926 (2020). https://doi.org/10.1093/ajh/hpaa074 Holton, K. F. in Glutamate and Neuropsychiatric Disorders: Current and Emerging Treatments (ed Zoran M. Pavlovic) 381-402 (Springer International Publishing, 2022). Pal, M. M. Glutamate: The Master Neurotransmitter and Its Implications in Chronic Stress and Mood Disorders. Front. Hum. Neurosci. 15 (2021). https://doi.org/10.3389/fnhum.2021.722323 Mazzitelli, M., Palazzo, E., Maione, S. & Neugebauer, V. Group II Metabotropic Glutamate Receptors: Role in Pain Mechanisms and Pain Modulation. Front. Mol. Neurosci. 11 (2018). https://doi.org/10.3389/fnmol.2018.00383 Brosnan, J. T. & Brosnan, M. E. Creatine: Endogenous Metabolite, Dietary, and Therapeutic Supplement. Annu. Rev. Nutr. 27 , 241-261 (2007). https://doi.org/https://doi.org/10.1146/annurev.nutr.27.061406.093621 Aroke, E. N. & Powell-Roach, K. L. The Metabolomics of Chronic Pain Conditions: A Systematic Review. Biol Res Nurs 22 , 458-471 (2020). https://doi.org/10.1177/1099800420941105 Teckchandani, S., Nagana Gowda, G. A., Raftery, D. & Curatolo, M. Metabolomics in chronic pain research. Eur J Pain 25 , 313-326 (2021). https://doi.org/10.1002/ejp.1677 Woolf, C. J. & Thompson, S. W. N. The induction and maintenance of central sensitization is dependent on N-methyl-d-aspartic acid receptor activation; implications for the treatment of post-injury pain hypersensitivity states. Pain 44 , 293-299 (1991). https://doi.org/https://doi.org/10.1016/0304-3959(91)90100-C Jang, J. H., Kim, D. W., Sang Nam, T., Se Paik, K. & Leem, J. W. Peripheral glutamate receptors contribute to mechanical hyperalgesia in a neuropathic pain model of the rat. Neuroscience 128 , 169-176 (2004). https://doi.org/https://doi.org/10.1016/j.neuroscience.2004.06.040 Cairns, B. E., Hu, J. W., Arendt-Nielsen, L., Sessle, B. J. & Svensson, P. Sex-Related Differences in Human Pain and Rat Afferent Discharge Evoked by Injection of Glutamate Into the Masseter Muscle. Journal of Neurophysiology 86 , 782-791 (2001). https://doi.org/10.1152/jn.2001.86.2.782 MacDonald, D. I., Wood, J. N. & Emery, E. C. Molecular mechanisms of cold pain. Neurobiology of Pain 7 , 100044 (2020). https://doi.org/https://doi.org/10.1016/j.ynpai.2020.100044 Story, G. M. et al. ANKTM1, a TRP-like channel expressed in nociceptive neurons, is activated by cold temperatures. Cell 112 , 819-829 (2003). https://doi.org/10.1016/s0092-8674(03)00158-2 Story, G. M. & Gereau, R. W. Numbing the senses: role of TRPA1 in mechanical and cold sensation. Neuron 50 , 177-180 (2006). https://doi.org/10.1016/j.neuron.2006.04.009 Sun, B., Bang, S.-I. & Jin, Y.-H. Transient receptor potential A1 increase glutamate release on brain stem neurons. NeuroReport 20 , 1002-1006 (2009). https://doi.org/10.1097/WNR.0b013e32832d2219 Sprague, J. E. & Arbeláez, A. M. Glucose counterregulatory responses to hypoglycemia. Pediatr Endocrinol Rev 9 , 463-473; quiz 474-465 (2011). Sapolsky, R. M. Stress and plasticity in the limbic system. Neurochem Res 28 , 1735-1742 (2003). https://doi.org/10.1023/a:1026021307833 Singh, K. D. et al. Standardization procedures for real-time breath analysis by secondary electrospray ionization high-resolution mass spectrometry. Anal Bioanal Chem 411 , 4883-4898 (2019). https://doi.org/10.1007/s00216-019-01764-8 Gisler, A. et al. An Interoperability Framework for Multicentric Breath Metabolomic Studies. iScience 25 , 105557 (2022). https://doi.org/https://dx.doi.org/10.2139/ssrn.4076338 García-Gómez, D. et al. Real-Time Quantification of Amino Acids in the Exhalome by Secondary Electrospray Ionization–Mass Spectrometry: A Proof-of-Principle Study. Clin. Chem. 62 , 1230-1237 (2016). https://doi.org/10.1373/clinchem.2016.256909 Tejero Rioseras, A. et al. Real-Time Monitoring of Tricarboxylic Acid Metabolites in Exhaled Breath. Anal. Chem. 90 , 6453-6460 (2018). https://doi.org/10.1021/acs.analchem.7b04600 Fontanillas, P., Kless, A., Team, a. R., Bothmer, J. & Tung, J. Y. Genome-wide association study of pain sensitivity assessed by questionnaire and the cold pressor test. PAIN 163 , 1763-1776 (2022). https://doi.org/10.1097/j.pain.0000000000002568 MacLachlan, C., Shipton, E. A. & Wells, J. E. The Cold Pressor Test as a Predictor of Prolonged Postoperative Pain, a Prospective Cohort Study. Pain and Therapy 5 , 203-213 (2016). https://doi.org/10.1007/s40122-016-0056-z Huang, M. et al. Early sympathetic neural responses during a cold pressor test linked to pain perception. Clinical Autonomic Research 31 , 215-224 (2021). https://doi.org/10.1007/s10286-019-00635-7 Keller, B. O., Sui, J., Young, A. B. & Whittal, R. M. Interferences and contaminants encountered in modern mass spectrometry. Analytica Chimica Acta 627 , 71-81 (2008). https://doi.org/https://doi.org/10.1016/j.aca.2008.04.043 Schlosser, A. & Volkmer-Engert, R. Volatile polydimethylcyclosiloxanes in the ambient laboratory air identified as source of extreme background signals in nanoelectrospray mass spectrometry. Journal of Mass Spectrometry 38 , 523-525 (2003). https://doi.org/https://doi.org/10.1002/jms.465 Zeng, J., Christen, A., Dev Singh, K., Frey, U. & Sinues, P. Comparison of Plasma Ionization- and Secondary Electrospray Ionization- High-resolution Mass Spectrometry for Real-time Breath Analysis. CHIMIA 76 , 127 (2022). https://doi.org/10.2533/chimia.2022.127 Liu, C. et al. Quantification of volatile organic compounds by secondary electrospray ionization-high resolution mass spectrometry. Analytica Chimica Acta 1180 , 338876 (2021). https://doi.org/https://doi.org/10.1016/j.aca.2021.338876 Nelson, L. S. The Shewhart Control Chart—Tests for Special Causes. Journal of Quality Technology 16 , 237-239 (1984). https://doi.org/10.1080/00224065.1984.11978921 Gisler, A. et al. An interoperability framework for multicentric breath metabolomic studies. iScience 25 , 105557 (2022). https://doi.org/10.1016/j.isci.2022.105557 Singh, K. D. G., Amanda; Osswald, Martin; Kohler, Malcolm; Van Den Anker, Johannes; Von Der Weid, Nicolas; Frey, Urs; Datta, Alexandre; Sinues, Pablo. European Patent EP4181781 - BREATH-BASED THERAPEUTIC DRUG MONITORING METHOD. (2021). Singh, K. D. S., Pablo. European Patent EP4370025 - METHOD FOR PROCESSING MASS SPECTROMETRY DATA OBTAINED FROM BREATH GAS. (2021). Mitchell, J. M. et al. New methods to identify high peak density artifacts in Fourier transform mass spectra and to mitigate their effects on high-throughput metabolomic data analysis. Metabolomics 14 , 125 (2018). https://doi.org/10.1007/s11306-018-1426-9 Gelman, A., Hill, J. & Vehtari, A. Regression and Other Stories . (Cambridge University Press, 2020). Storey, J. D. A direct approach to false discovery rates. J. Roy. Stat. Soc. Ser. B. (Stat. Method.) 64 , 479-498 (2002). https://doi.org/10.1111/1467-9868.00346 Kind, T. & Fiehn, O. Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry. BMC Bioinformatics 8 , 105 (2007). https://doi.org/10.1186/1471-2105-8-105 Gustavsen, J. A., Pai, S., Isserlin, R., Demchak, B. & Pico, A. R. RCy3: Network biology using Cytoscape from within R. F1000Res 8 , 1774 (2019). https://doi.org/10.12688/f1000research.20887.3 Additional Declarations Yes there is potential Competing Interest. PS is co-founder and board member of Deep Breath Intelligence (DBI) AG, a company that provides services in the field of breath analysis. Kapil Dev Singh is partially employed by the same company. UKBB is a shareholder of said company. Supplementary Files PROTOCOLV1.2.pdf Protocol InfoIC1V1.1.pdf Info and IC supplementary.docx Fig. S1, Fig. S2, Fig. S3, S Table 3, Inclusion Exclusion Checklist SupplementaryTable1.xlsx Supplementary Table1 SupplementaryTable2.xlsx Supplementary Table2 Cite Share Download PDF Status: Published Journal Publication published 21 Apr, 2026 Read the published version in iScience → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6048423","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":422190328,"identity":"b9cb3ee9-d53f-4105-b289-729f10b6d406","order_by":0,"name":"Mélina Richard","email":"","orcid":"https://orcid.org/0000-0003-4432-7905","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Mélina","middleName":"","lastName":"Richard","suffix":""},{"id":422190329,"identity":"7bfb7e1d-34f8-4587-82e4-f4f83a9c3e00","order_by":1,"name":"Kapil Singh","email":"","orcid":"","institution":"University Children’s Hospital Basel","correspondingAuthor":false,"prefix":"","firstName":"Kapil","middleName":"","lastName":"Singh","suffix":""},{"id":422190330,"identity":"4c17984f-7ff0-4138-a9dd-ba115f75f770","order_by":2,"name":"Dilan Sezer","email":"","orcid":"https://orcid.org/0000-0003-4751-8387","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Dilan","middleName":"","lastName":"Sezer","suffix":""},{"id":422190331,"identity":"6af7ffa5-142d-4d2d-97a3-03cfe899598c","order_by":3,"name":"Sarah Buergler","email":"","orcid":"","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Sarah","middleName":"","lastName":"Buergler","suffix":""},{"id":422190332,"identity":"14e7b9c1-112d-461d-8b2d-989fb020e4e0","order_by":4,"name":"Luana Palermo","email":"","orcid":"https://orcid.org/0009-0007-0381-3962","institution":"University of Bern","correspondingAuthor":false,"prefix":"","firstName":"Luana","middleName":"","lastName":"Palermo","suffix":""},{"id":422190333,"identity":"0ad02b87-83cb-4332-91d9-59e4242205e4","order_by":5,"name":"Yannick Schulz","email":"","orcid":"","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Yannick","middleName":"","lastName":"Schulz","suffix":""},{"id":422190334,"identity":"e8443d14-ee73-4853-9ec7-150e73f4f0ad","order_by":6,"name":"Zhifeng Tang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Zhifeng","middleName":"","lastName":"Tang","suffix":""},{"id":422190335,"identity":"9ef1c0d9-d23f-430c-b6a1-8ab6045b2543","order_by":7,"name":"Xin Luo","email":"","orcid":"","institution":"College of Environment and Climate, Institute of Mass Spectrometry and Atmospheric Environment, Guangdong Provincial Key Laboratory of Speed Capability Research, Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Luo","suffix":""},{"id":422190336,"identity":"c6975f13-7a9d-45de-9ab0-ec5cc1449f53","order_by":8,"name":"Urs Frey","email":"","orcid":"https://orcid.org/0000-0003-3773-2822","institution":"University Children's Hospital (UKBB), University of Basel, Basel","correspondingAuthor":false,"prefix":"","firstName":"Urs","middleName":"","lastName":"Frey","suffix":""},{"id":422190337,"identity":"039d9898-a525-4832-b2be-97f0acb28d89","order_by":9,"name":"Philippe Cattin","email":"","orcid":"https://orcid.org/0000-0001-8785-2713","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Philippe","middleName":"","lastName":"Cattin","suffix":""},{"id":422190338,"identity":"0be0b0f3-c546-4f1b-91df-9a4cfdd42a94","order_by":10,"name":"Xue Li","email":"","orcid":"","institution":"College of Environment and Climate, Institute of Mass Spectrometry and Atmospheric Environment, Guangdong Provincial Key Laboratory of Speed Capability Research, Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Xue","middleName":"","lastName":"Li","suffix":""},{"id":422190339,"identity":"d3bbaad5-dffd-4bc0-b77b-e1b2fe43ed7e","order_by":11,"name":"Jens Gaab","email":"","orcid":"","institution":"University of Basel","correspondingAuthor":false,"prefix":"","firstName":"Jens","middleName":"","lastName":"Gaab","suffix":""},{"id":422190327,"identity":"fff4a337-ee1c-4c10-8e77-681c4bf4ca28","order_by":12,"name":"Pablo Sinues","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIie3PMQrCMBSA4YQOLgFXpT3EmwwZxKsIhbq0uolTcepU9x7FMSWDS9W1g0NFqItIoYtuJh1UkEbcHPIP4RHykQQhk+kvI5irtZkrOYAa8FJL0JPg5BfSZCn4ldDOlvMKhQ7tuaIeRgeHdjYFuq/bCYun4zRBgrDE8+wgKgmLfcCrrJ0A90HItxHI/YEVRILIHWThSEP2F0VCSWZ1zRTZnws9yZtbLHULsrEi+Ri0hCVX+RepICsH/XinPnWBdKUhtBukRbUIR7BxT9VtLka0OzkWd93D3tZXvB18HjaZTCbTZw/OqVO7lN3i5wAAAABJRU5ErkJggg==","orcid":"","institution":"University Children’s Hospital Basel","correspondingAuthor":true,"prefix":"","firstName":"Pablo","middleName":"","lastName":"Sinues","suffix":""}],"badges":[],"createdAt":"2025-02-17 13:35:29","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":true,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6048423/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6048423/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1016/j.isci.2026.115857","type":"published","date":"2026-04-22T00:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":77685623,"identity":"b2bc00d6-59d0-4ab0-a954-c1b439c4f553","added_by":"auto","created_at":"2025-03-04 09:02:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":247844,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlow chart of the overall study design and data analysis pipeline to unveil pain-induced metabolic shifts. a \u003c/strong\u003eN = 40 participants from two centers (i.e., discovery cohort n=19, and validation cohort n=21) were involved in this study. They underwent the same CPT intervention to induce pain and subsequent breath metabolic fingerprinting procedures. Breath specimens were subjected to real-time mass spectrometry-based metabolomics analysis before and after pain induction. \u003cstrong\u003eb\u003c/strong\u003eIt followed a univariate and multivariate analysis pipeline to identify significant changes in exhaled metabolites. \u003cstrong\u003ec\u003c/strong\u003e Correlation and pathway analysis were then used to evaluate the similarities in response to CPT between both cohorts and for biological interpretation. \u003cstrong\u003ed \u003c/strong\u003eNeural network models were developed to ultimately predict whether the exhaled metabolic signature corresponded to pain/no-pain.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/c1b5be598e7da6f693d5de6b.png"},{"id":77683828,"identity":"8136287b-e6be-4d8b-863f-235caeb48b3f","added_by":"auto","created_at":"2025-03-04 08:54:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":331057,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConcentrations of exhaled metabolites are increased upon pain induction. Partial\u003c/strong\u003e Least Squares (PLS) score plots for both discovery and validation datasets show clear separation between pre-CPT (blue circles) and post-CPT (red crosses) metabolic profiles. This clustering suggests significant metabolic shifts induced by the CPT intervention, with consistent patterns across both datasets, supporting the robustness of the observed changes. \u003cstrong\u003eb\u003c/strong\u003e Volcano plot analysis performed on after- vs. post-CPT of all 5,058 metabolic features for the discovery and validation data sets. P-values were determined by a two-sided paired t-test followed by adjustment for multiple comparison. Differentially regulated metabolic features detected in just one data set are indicated by red dots. Green dots represent overlapping features between discovery and validation cohorts. \u003cstrong\u003ec\u003c/strong\u003e Representative example of an overlapping mass spectral identified by the volcano plot analysis, which was mapped to arginine. Each spectrum corresponds to one participant, whereas the dashed line indicates the theoretical m/z of protonated arginine ± 1 ppm (solid vertical lines). \u003cstrong\u003ed\u003c/strong\u003e Signal intensity (Log10(AUC), see methods) of the same feature depicted as a box plot overlaid with points, whereas every line represents an individual participant connecting pre-(blue) and post-CPT (red) boxplots. Discovery: mean Log\u003csub\u003e2\u003c/sub\u003eFC = 1.90; FDR \u0026lt; 0.001 and validation: mean Log\u003csub\u003e2\u003c/sub\u003eFC = 1.58; FDR \u0026lt; 0.001.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/70b7d85acf582ad7937fce6e.png"},{"id":77685622,"identity":"382208ad-ce83-4886-8115-0d9904b25556","added_by":"auto","created_at":"2025-03-04 09:02:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":348389,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-cohort correlation analysis of metabolite networks reveals consistent behavior following CPT-induced pain. a\u003c/strong\u003e Correlation networks for discovery and validation cohorts at a ρ cutoff of 0.7. Node colors represent mean Log\u003csub\u003e2\u003c/sub\u003e fold change (Log\u003csub\u003e2\u003c/sub\u003eFC), with higher fold changes clustering at the network periphery, indicating consistent intervention-driven effects across cohorts. \u003cstrong\u003eb\u003c/strong\u003e Cross-site similarity metrics reveal high structural similarity between networks, as indicated by significant Trace-Based Distance (0.678, p \u0026lt; 0.001) and eigenvalue correlation (0.992, p \u0026lt; 0.001). Metrics such as Jaccard Index and Edge Differences highlight increasing divergence at higher ρ cutoffs, reflecting site-specific variations in weaker metabolic connections. \u003cstrong\u003ec\u003c/strong\u003e Network topology analysis shows reduced clustering and increased fragmentation at higher ρ cutoffs, capturing the strongest intervention-driven pathways. Degree-FC correlations remain consistently negative, with hub nodes showing smaller fold changes. Metrics such as path length, diameter, and betweenness centrality highlight subtle differences in network topology while underscoring conserved metabolic responses across cohorts.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/43b7cb7a69a5f2885bedc1fa.png"},{"id":77683832,"identity":"f8146022-4e81-4c2d-ae6f-b27e34492e37","added_by":"auto","created_at":"2025-03-04 08:54:21","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":295637,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePathway-based degree and enrichment analysis reveals a consistent multifaceted metabolic cascade in both cohorts. a\u003c/strong\u003e Comparison of median Log\u003csub\u003e2\u003c/sub\u003e fold changes (Log\u003csub\u003e2\u003c/sub\u003eFC) between discovery and validation cohorts across metabolic pathways shows strong agreement. \u003cstrong\u003eb\u003c/strong\u003e Degree metrics, reflecting network connectivity, also exhibit high concordance between cohorts. \u003cstrong\u003ec\u003c/strong\u003e Hierarchical clustering of pathways based on intra-pathway Log\u003csub\u003e2\u003c/sub\u003eFC and degree correlations highlights consistent perturbations in pathways such as alpha-linolenic acid metabolism (ALA), fructose and mannose metabolism (FMM), ketone body metabolism (KBM), butyrate metabolism (BUT), arginine and proline metabolism (APM), cysteine metabolism (CYS) and histidine metabolism (HIS). These pathways underscore the body's metabolic response to CPT, focusing on energy homeostasis and biosynthesis. \u003cstrong\u003ed\u003c/strong\u003e A schematic representation of enriched pathways, derived from mummichog analysis, connects Aminoacyl-tRNA biosynthesis, Cysteine and methionine metabolism, Butanoate metabolism, Alanine, aspartate, and glutamate metabolism, and Arginine and proline metabolism. These pathways integrate fatty acid, carbohydrate, and amino acid metabolism to reflect the primary metabolic shifts following CPT.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/517db5c2c722194bfdd2b1c2.png"},{"id":77683821,"identity":"3b771b47-a547-4474-a406-573dfe532fff","added_by":"auto","created_at":"2025-03-04 08:54:20","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":88088,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNeural network-based prediction of pain state using breath mass spectral fingerprints. a\u003c/strong\u003e Receiver operating characteristic (ROC) curve for the neural network model trained to classify pre- and post-CPT breath fingerprints. The model achieved an area under the curve (AUC) of 0.86 and an overall accuracy of 78%. \u003cstrong\u003eb\u003c/strong\u003eHeatmap of classification probabilities for individual breath samples across discovery and validation cohorts. Hierarchical clustering revealed three distinct clusters: (1) a \"no-pain\" cluster consisting of samples primarily classified with low probabilities of pain, (2) a \"pain\" cluster of post-CPT samples with high probabilities of correct classification, and (3) a mixed cluster where some pre-CPT samples were incorrectly classified as “pain,” potentially reflecting a heightened metabolic response prior to the intervention.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/dffe13d534f6f6afb4413489.png"},{"id":108805252,"identity":"fee86dac-79ea-4534-8b32-973c771c248c","added_by":"auto","created_at":"2026-05-08 15:25:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1523350,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/752b4eaa-fe32-403d-bed9-ec5870f8ca1c.pdf"},{"id":77685626,"identity":"582e4613-9392-49c5-b828-31810c3778ff","added_by":"auto","created_at":"2025-03-04 09:02:20","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":503488,"visible":true,"origin":"","legend":"Protocol","description":"","filename":"PROTOCOLV1.2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/a4d869def9cd4a26f9c62bc7.pdf"},{"id":77689414,"identity":"fb8175fa-dcc9-42d7-8565-19c91b810889","added_by":"auto","created_at":"2025-03-04 09:26:20","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":323978,"visible":true,"origin":"","legend":"Info and IC","description":"","filename":"InfoIC1V1.1.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/8ceb23fda8d6dbd397dacc48.pdf"},{"id":77683822,"identity":"909c3796-e400-479f-8bd6-264d0cdd5fc0","added_by":"auto","created_at":"2025-03-04 08:54:20","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1207871,"visible":true,"origin":"","legend":"Fig. S1, Fig. S2, Fig. S3, S Table 3, Inclusion Exclusion Checklist","description":"","filename":"supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/10a6cb29d8814ddea40e527d.docx"},{"id":77683829,"identity":"f2f083cd-5412-4b09-89a9-1ffcdabcb451","added_by":"auto","created_at":"2025-03-04 08:54:21","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":970935,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Table1\u003c/p\u003e","description":"","filename":"SupplementaryTable1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/f1b6e563be667145985ba4ab.xlsx"},{"id":77687367,"identity":"1b235734-dd76-46d8-a0d5-c9a73ecf53ab","added_by":"auto","created_at":"2025-03-04 09:10:20","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":17189,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Table2\u003c/p\u003e","description":"","filename":"SupplementaryTable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6048423/v1/ef016b4a763e54582c857ba7.xlsx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nPS is co-founder and board member of Deep Breath Intelligence (DBI) AG, a company that provides services in the field of breath analysis. Kapil Dev Singh is partially employed by the same company. UKBB is a shareholder of said company.","formattedTitle":"Pain induces a rapid characteristic metabolic signature detectable in breath","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePain, in its broad range of severities, remains to be a major human and healthcare systems\u0026rsquo; burden\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Pain is the most frequent cause of admissions to emergency facilities and is also commonly experienced in primary healthcare settings\u003csup\u003e\u003cspan additionalcitationids=\"CR4\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e and significantly contributes to health care costs. Straightforward strategies for lowering costs in healthcare is acute pain management, and several nations have published recommendations or evidence syntheses on pain management\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. Nevertheless, pain assessment is often based on observer dependent measures of pain and related scoring systems.\u003c/p\u003e \u003cp\u003eDespite these efforts, the objectification of pain presents a significant clinical challenge particularly in vulnerable risk groups. This is particularly true in children\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e, elderly individuals\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, patients with cognitive impairment\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e and unconscious patients\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. It is critically important to accurately assess pain in these populations due to the heightened risk of undertreatment. E.g. pain assessment in children depends on factors such as cognitive development, clinical context, and pain typology, with self-reporting used for children over 6 years and behavioral scales for younger children\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. As an example of objective pain assessment, studies have explored skin conductance (SC) as a potential tool for measuring acute pain. While some findings suggest SC may be a useful indicator, others have reported inconsistent results \u003csup\u003e\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. In patients with cognitive impairment, the undertreatment of pain remains a significant issue in healthcare.\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e In elderly patients, healthcare professionals tend to underestimate pain needs and under-prescribe medications due to lack of objective measures\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Even more challenging is the pain assessment and management in unconscious ICU patients due to communication barriers. Unrecognized or undertreated pain affects 70% of ICU patients, leading to serious physical and psychological consequences\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Furthermore, in the operating theatre, there is a need for rapid assessment of changes in pain during a surgical intervention or in the recovery room\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn the context of personalized health, we lack the necessary knowledge to tailor evidence-based approaches for individual patients or even specific subclasses. Most human research is grounded in population-based outcomes, yet the reality is that many patients respond to interventions in ways that differ significantly from the average\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. Hence it is important to investigate further into underlying molecular mechanisms associated with pain responses.\u003c/p\u003e \u003cp\u003eWe hypothesize that characterizing the exhaled metabolome is ideally suited for these four challenges such as \u003cem\u003ethe need for an objective observer independent pain metrics\u003c/em\u003e, the \u003cem\u003esuitability for vulnerable risk groups\u003c/em\u003e, \u003cem\u003ethe possibility to detect immediate changes in pain during interventions\u003c/em\u003e and the \u003cem\u003eassessment of metabolic response profiles enabling potentially personalized treatment\u003c/em\u003e regimes in the future). Exhalomics offers the possibility to detect rapid \u0026mdash;in the time scales of minutes\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e\u0026mdash;fluctuations in concentrations as a result of a stressor such as acute pain. Metabolites have also been demonstrated to have exceptional predictive abilities and to closely mirror the actual phenotype\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, providing an important path towards personalized medicine\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e, including pain \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Among the whole metabolome, the most volatile subset of it is thought to play a key pain-signaling role in the animal kingdom: olfactory cues may convey pain to others\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn this proof-of-concept human experimental study in two cohorts, we propose a real-time, non-invasive assessment of biochemical changes of experimentally induced pain by cold pressor test (CPT) by harnessing the metabolome via exhaled breath. The CPT is the most popular pain-provocation test in history\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. It entails keeping one hand submerged in ice water until it becomes intolerable, at which time the hand is removed\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. The objective of this study was to assess whether there is a CPT-induced metabolic signature that may allow us to gain further insights into the molecular mechanisms of pain. To this end, we applied a well-established, non-invasive and real-time breath metabolomics pipeline first in a discovery cohort, and subsequently in a validation cohort.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design\u003c/h2\u003e \u003cp\u003eA total of n\u0026thinsp;=\u0026thinsp;20 healthy participants were enrolled in the University Children\u0026rsquo;s Hospital Basel in Switzerland (\u003cem\u003ei.e.\u003c/em\u003e, discovery cohort). One participant withdrew the consent retrospectively, hence n\u0026thinsp;=\u0026thinsp;19 participants\u0026rsquo; data were analyzed. A validation cohort of n\u0026thinsp;=\u0026thinsp;21 healthy participants were recruited within Jinan University in China (\u003cem\u003ei.e.\u003c/em\u003e, validation cohort; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The demographic characteristics of the discovery and validation cohorts were compared. The median age of participants in the discovery cohort was 26.0 years (interquartile range [IQR]: 5.5 years), while the median age in the validation cohort was 24.0 years (IQR: 5.0 years). A Mann-Whitney U test revealed no significant difference in age distribution between the two cohorts (U\u0026thinsp;=\u0026thinsp;187.0, p\u0026thinsp;=\u0026thinsp;0.744). Regarding gender distribution, the discovery cohort included 7 males and 12 females, while the validation cohort comprised 12 males and 9 females. A chi-square test found no significant difference in gender distribution between the cohorts (χ\u0026sup2; = 0.935, p\u0026thinsp;=\u0026thinsp;0.334). The N\u0026thinsp;=\u0026thinsp;40 participants of both sites completed four exhalation measurements, two before- (-15 and \u0026minus;\u0026thinsp;5 min) and two after-CPT (+\u0026thinsp;0 and +\u0026thinsp;25 min) intervention. A multifaceted univariate and multivariate data analysis pipeline was deployed to identify altered metabolites/metabolic pathways because of induced pain, as well as to predict whether the breath mass spectral fingerprint corresponds to a pre- or post-CPT sample.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePain-induced physiological and metabolic alterations\u003c/h3\u003e\n\u003cp\u003eThe pain threshold, as assessed by the time the participants could withstand their hand in cold water, varied substantially across individuals, as reported in literature\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. The median (interquartile range) withstanding the hand in iced water was 66 (200) and 36 (24) s for the discovery and validation cohort, respectively. Such intervention resulted in a significant (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) increase in blood pressure (\u003cb\u003eSupplementary Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e)\u003c/b\u003e. A total of 5,058 mass spectral features were detected in both cohorts, of which 1,377 could be mapped to at least one metabolite from the Human Metabolome Database\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e (\u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e). Whether pain induction came accompanied by a sizeable change in the overall exhaled metabolic profile was initially assessed by partial least squared-discriminant analysis (PLS-DA). A 10-fold cross-validated analysis revealed a robust classification performance for both cohorts, as assessed by the accuracy (discovery: 0.97; validation: 0.93) and discriminant Q\u003csup\u003e2 30\u003c/sup\u003e(discovery: 0.74; validation: 0.84). Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea shows the resulting score plot based on SESI-HRMS breath mass spectra from the participants pre- and post-CPT. A clear shift of the data points upon the intervention is observed in both the discovery and validation cohorts, suggesting that CPT induces a consistent metabolic shift. To gain further insights into the strength and significance of such changes at the molecular level, we conducted a volcano plot analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The figure indeed displays a striking change in the metabolic profile after pain induction, whereby the signal intensity of most of the features was increased upon intervention, as evidenced by the asymmetric volcano plot. We compared the area under the curve of both measurements before pain against the area under the curve of both measurements after pain (i.e. measurements 1\u0026thinsp;+\u0026thinsp;2 versus 3\u0026thinsp;+\u0026thinsp;4). A total of 1,153 features were found to be significantly upregulated (Log\u003csub\u003e2\u003c/sub\u003eFC\u0026thinsp;\u0026ge;\u0026thinsp;1.5 \u0026amp; FDR\u0026thinsp;\u0026le;\u0026thinsp;0.01) in the discovery cohort. In contrast, just 18 features were significantly decreased (Log\u003csub\u003e2\u003c/sub\u003eFC \u0026le; -1.5 \u0026amp; FDR\u0026thinsp;\u0026le;\u0026thinsp;0.01) after pain induction. The analysis of the validation cohort revealed a remarkably similar picture, which exhibited 790 upregulated features and 13 downregulated features, respectively. Interestingly, 416 upregulated and 6 downregulated features overlapped between the two study sites. The mass spectra for all participants of one such example, which was mapped to arginine, is displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec. It shows a generalized and significant increase as a response to the CPT across all subjects for the discovery (mean Log\u003csub\u003e2\u003c/sub\u003eFC\u0026thinsp;=\u0026thinsp;1.90; FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.001 ) and validation (mean Log\u003csub\u003e2\u003c/sub\u003eFC\u0026thinsp;=\u0026thinsp;1.58; FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.001) cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed\u003cb\u003e)\u003c/b\u003e. \u003cb\u003eSupplementary Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/b\u003e shows additional examples exhibiting a similar trend for creatine, threonine and asparagine.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo gain further insights into whether the relationships across the metabolic features were conserved in the discovery and validation cohorts, we conducted a correlation analysis. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea shows the resulting correlation network (ρ cut-off \u0026gt;\u0026thinsp;0.7 in both sites). The color coding of the nodes signifies the mean fold-change (FC) on Log\u003csub\u003e2\u003c/sub\u003e scale (i.e. Log\u003csub\u003e2\u003c/sub\u003eFC).. It becomes apparent that the features experiencing the strongest response after pain induction tend to cluster together and are in the periphery of the network, whereas those with the lowest FC tend to also correlate and are at the core of the network. A visual inspection of the color distribution across both networks provides a sense of resemblance, indicating that also at that level, the data structure is conserved in the validation set.\u003c/p\u003e \u003cp\u003eFurther analysis of discovery and validation networks across various ρ cutoffs reveals significant structural similarity between the two datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). The correlation matrix distance\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e of 0.678 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) indicates that the networks are highly similar. An eigenvalue correlation of 0.992 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) confirms that the overall variance structures of the two networks are almost identical. However, the moderate correlation of correlations (0.187, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and the Jaccard Index, which decreases from 0.59 at ρ\u0026thinsp;=\u0026thinsp;0.1 to 0.06 at ρ\u0026thinsp;=\u0026thinsp;0.90, suggest that differences in edge patterns and connectivity become more pronounced as weaker correlations are pruned. The weighted edge correlation, ranging from 0.22 to 0.31, reflects moderate consistency in edge weights, further underscoring a partial overlap in network connectivity. The edge difference metric, which decreases steadily from 0.305 at ρ\u0026thinsp;=\u0026thinsp;0.1 to 0.023 at ρ\u0026thinsp;=\u0026thinsp;0.9, highlights the reduction in the magnitude of edge weight differences between the two networks as weaker correlations are filtered out. This indicates a progressive alignment in the remaining stronger connections. Degree-FC correlation is consistently negative, becoming more pronounced at higher ρ cutoffs, with values ranging from \u0026minus;\u0026thinsp;0.074 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) to -0.113 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in the discovery cohort and from \u0026minus;\u0026thinsp;0.220 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) to -0.250 (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in the validation cohort. This suggests that higher-degree nodes tend to exhibit smaller fold changes, reinforcing the role of hub nodes in maintaining network stability despite local differences. Both networks remain fully connected up to a ρ cutoff of 0.6, beyond which the number of connected components increases rapidly, indicating network fragmentation. The diameter of the networks remains stable at lower ρ values, with a value of 2, but begins to increase as the networks become sparser, reflecting reduced global connectivity. Metrics such as the clustering coefficient, which decreases from 0.38 to 0.05, and the average path length, which increases from 1.27 to 5.56, remain stable at lower ρ cutoffs but diverge as ρ increases. This divergence indicates that local connectivity and efficiency diminish as weaker edges are pruned, further contributing to network sparsity. Betweenness centrality shows a moderate decrease, from 679 to 480, as ρ increases, suggesting a shift in the importance of key nodes in maintaining overall connectivity. This reflects changes in the centrality landscape of the networks as they become fragmented. Overall, while local connectivity and specific edge patterns vary significantly with increasing ρ cutoffs, the global structural similarity between the discovery and validation networks suggests that the networks are underpinned by similar underlying metabolic processes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003ePain-induced altered metabolic pathways\u003c/h3\u003e\n\u003cp\u003eThe univariate, multivariate and correlation network data analysis approaches presented above suggest a stark and rapid metabolic perturbation as a result of the CPT intervention consistent in both cohorts. To generate further hypothesis into the specific altered regions of human metabolome, we conducted a pathway-based degree analysis. A total of 55 metabolic pathways were associated with the 5,058 mass spectral features overall detected in both cohorts (\u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e). To increase the robustness of the analysis, we further considered only those pathways represented by at least five mass spectral peaks, reducing the list to 32 pathways. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea and Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb show an excellent overall agreement for the median Log\u003csub\u003e2\u003c/sub\u003eFC and degree between the discovery and validation cohorts. Further insights into the intra-pathway correlation for the Log\u003csub\u003e2\u003c/sub\u003eFC and the degree could be gained by conducting a hierarchical cluster analysis Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec. It emerges from this analysis that the most consistent behavior between the discovery and validation cohorts for both, the Log\u003csub\u003e2\u003c/sub\u003eFC and the degree, was found for alpha-linolenic acid metabolism (ALA), fructose and mannose metabolism (FMM), ketone body metabolism (KBM), butyrate metabolism (BUT), arginine and proline metabolism (APM), aspartate metabolism (ASP), cysteine metabolism (CYS), and histidine metabolism (HIS). These pathways collectively highlight the body\u0026rsquo;s ability to convert and utilize fatty acids, carbohydrates, and amino acids for energy homeostasis and biosynthetic needs\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo complete the biological interpretation of CPT-driven metabolic shifts, we conducted an enrichment analysis using the mummichog algorithm\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. We considered as significant feature list those that were significantly altered in both study sites (i.e., 8% of the total feature list; 416/6 upregulated/downregulated; \u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e\u003c/b\u003e). The top five pathways (p\u003csub\u003egamma\u003c/sub\u003e \u0026lt; 0.1) include Aminoacyl-tRNA biosynthesis, Cysteine and methionine metabolism, Butanoate metabolism, Alanine, aspartate and glutamate metabolism and Arginine and proline metabolism (\u003cb\u003eSupplementary Fig. S3\u003c/b\u003e), comprising a total of 58 compounds mapped to these pathways (\u003cb\u003eSupplementary Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/b\u003e). Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed shows a schematic representation of the how the main identified pathways are interconnected to provide an overview of the main metabolic alterations driven by CPT.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eNeural networks predict whether a breath mass spectral fingerprint is associated with pain\u003c/h3\u003e\n\u003cp\u003eThe data presented above suggests rapid and stark metabolic changes taking place as a result of CPT intervention. Therefore, we attempted to establish a neural network model to test the hypothesis of whether a first-line breath test could determine whether a patient is in pain. The resulting ROC curve (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea) summarizes the prediction performance whereby the AUC\u0026thinsp;=\u0026thinsp;0.856 and the overall model accuracy was 78%. When analyzing the results at an individual level (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb), we observe that the majority of the participants were correctly classified whereby the first two measurements (\u003cem\u003ei.e.\u003c/em\u003e, pre-CPT) were assigned to belong to the \u0026ldquo;pain\u0026rdquo; mass spectral signature with a low probability and the last two measurements (\u003cem\u003ei.e.\u003c/em\u003e, post-CPT) with a high probability. However, some individuals were systematically misclassified whereby the second measurement was misclassified as \u0026ldquo;pain\u0026rdquo; and the third measurement as \u0026ldquo;no-pain\u0026rdquo;. The hierarchical clustering analysis revealed three main clusters. The first cluster predominantly consisted of samples classified as \"no-pain,\" including most pre-CPT measurements, suggesting that their metabolic response to pain induction was either delayed or less pronounced. The second cluster comprised post-CPT samples that were consistently classified as \"pain,\" reflecting a robust metabolic shift induced by the CPT intervention. The third cluster contained a mix of pre- and post-CPT samples, where some pre-CPT samples were misclassified as \"pain.\" This misclassification may point to individual variability in baseline metabolic states, potentially influenced by pre-procedural nervousness or stress, as the CPT intervention might have been anticipated. Overall, the correct classification rate was 92.5%, 70%, 70% and 87.5% for the first, second, third and fourth breath measurement, respectively. Such misclassifications were observed both in the discovery and the validation cohorts.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this proof-of-concept study, we presented and validated a novel observer independent method which can reliably detect pain in human adults with an accuracy of 78% and an area under the curve of 0.86. The study was not designed to directly compare this novel method to standard clinical pain scores such as Numeric Rating Scale (NRS), Visual Analog Scale (VAS) or Verbal Rating/Descriptor Scale (VRS/VDS). It moreover offers and validates a novel technology which must be tested in future studies in various patient groups under various clinical conditions. Nevertheless, the novel observer independent technology offers the possibility to be applicable to vulnerable risk groups of patients such as infants, children, the elderly, patients with cognitive impairment or sedated or unconscious patients in ICU and operating theatre. In the latter setting, the possibility to detect immediate changes in pain during interventions makes the techniques potentially extremely useful to guide intervention and analgetic regimen during anesthesia.\u003c/p\u003e \u003cp\u003eFurthermore, the assessment of metabolic response profiles enabling potentially personalized treatment. Pain is a highly complex sensation which is objectively difficult to describe and likely has a high individual characteristic dependent on disease and circumstances. Research on skin conductance (SC) as a tool for assessing acute pain in infants and children has yielded mixed results. While some studies found SC to be a potentially useful indicator of pain, others reported inconsistent findings. Hullett et al. demonstrated that SC could accurately predict the absence of moderate to severe pain in postoperative pediatric patients \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Similarly, Dalal et al. found that peak SC values may serve as indicators of unmitigated pain in infants \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. However, Solana et al. \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e concluded that SC was not more sensitive or faster than clinical scales for pain assessment in critically ill children. Hu et al. reported inconsistent validity evidence for SC in pain assessment, noting significant correlations with unidimensional behavioral pain scales but not with multidimensional measurements \u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Overall, while SC shows promise as a pain assessment tool, further research is needed to establish its diagnostic accuracy and clinical applicability across different pediatric populations and pain contexts.\u003c/p\u003e \u003cp\u003eThe International Association for the Study of Pain has defined pain as “an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage”\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Characterizing such a sensory and emotional experience at a molecular level is an inherently challenging task. This pain response involves the activation of the sympathetic nervous system (SNS), which increases sympathetic tone across various organs. The intensity of this pain-related stress response is proportional to the severity of the injury and the level of pain experienced. The body's reaction to pain serves to mobilize energy resources and prioritize vital functions needed for survival. We showed that the bodies response to pain results in changes in metabolism expressed in the changes in exhaled metabolic profiles. These profiles may inherit properties of a patient specific pain response and could potentially help to guide future treatment strategies. These exhaled metabolic profiles were assessed in a non-invasive manner as follows: We used CPT as a proxy for pain induction combined with a well-established real-time breath metabolomics pipeline, applying it in both a European (discovery) and a Chinese (validation) cohort. Our results suggest that pain, as induced by CPT, triggers within 15 min a significant upregulation of ~ 400 metabolic mass spectral features, highly conserved across these populations. Subsequent enrichment and pathway-degree analysis flagged coordinated metabolic adjustments in aminoacyl-tRNA biosynthesis, cysteine/methionine metabolism, butanoate metabolism, alanine/aspartate/glutamate metabolism, and arginine/proline metabolism. Enhanced aminoacyl-tRNA biosynthesis likely supports stress-induced protein synthesis demands, as aminoacyl-tRNA synthetases and related factors can be upregulated under cellular stress\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. Meanwhile, the upregulation of cysteine and methionine metabolism can reflect both increased antioxidant defenses and shifts in methylation crucial for redox control and possibly catecholamine turnover\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e. Butanoate metabolism may help supply alternative energy substrates and modulate inflammation during acute stress, supported by evidence that short-chain fatty acids affect host immune and metabolic homeostasis\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. In parallel, alterations in alanine, aspartate, and glutamate metabolism—particularly glutamate’s central role in nociceptive transmission—could reinforce pain signaling in the central nervous system\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. Finally, the arginine–proline axis, which governs nitric oxide (NO) production\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e and tissue remodeling, may synergize with sympathetic activation to fine-tune vascular tone, culminating in the hallmark blood pressure increase during CPT.\u003c/p\u003e \u003cp\u003eThis high-level metabolic shift aligns with prior work showing CPT-induced alterations at both transcriptome and metabolome levels, reflecting SNS activation in stress and pain responses\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. For example, arginine was previously found to correlate with gene expression changes following CPT, consistent with our observation of elevated arginine (average Log\u003csub\u003e2\u003c/sub\u003e(FC) = 1.90 in the discovery cohort; 1.58 in the validation cohort; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec and Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed), a key amino acid in NO production and vascular regulation. Additionally, arginine plays a role in energy mobilization, supporting the body's efforts to prioritize survival functions when the SNS is activated, especially during stress and pain\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. Glutamic acid also increased (average Log\u003csub\u003e2\u003c/sub\u003e (FC) = 1.74 in discovery; 1.44 in validation) and, as an excitatory neurotransmitter, may intensify nociceptive signaling while impacting blood pressure regulation\u003csup\u003e\u003cspan additionalcitationids=\"CR46\" citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e–\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. Gamma-aminobutyric acid (GABA), in contrast, generally acts inhibitory but showed elevated levels as well (average Log\u003csub\u003e2\u003c/sub\u003e (FC) = 1.53 in discovery; 1.37 in validation). Dysregulation in GABA balance can contribute to heightened SNS excitability, pain perception, and blood pressure. Creatine (average Log\u003csub\u003e2\u003c/sub\u003e (FC) = 1.62 in discovery; 1.92 in validation; \u003cb\u003eSupplementary Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/b\u003e) further supports energy metabolism during heightened SNS activation\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eInterestingly, these findings parallel meta-analyses on chronic pain that report alterations in amino acids (e.g., glutamine, serine, phenylalanine) and macromolecule metabolism intermediates (e.g., succinate, citrate, acetylcarnitine, and N-acetylornithine)\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e. When we look further into cold induced pain, a recent meta-analysis on metabolomics in chronic pain research \u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e has found, that pathways resulting in increased glutamate levels, either systemically or locally, have been exhibited in a variety of conditions, including fibromyalgia, Complex regional pain syndrome, trapezius myalgia, and mixed musculoskeletal pain conditions. It is widely recognized that glutamate signaling plays a part in both central and peripheral sensitization\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e,\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. Although the type of pain in our study was not chronic, this constancy holds true for many types of pain. Furthermore, the sensitivity to decreasing temperature is significantly increased among chronic pain patients throughout conditions\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e. Glutamate metabolites may serve as indicators of nociception in a variety of pain disorders that also show in cold receptors potential where cold temperatures and irritating substances can stimulate the transient receptor potential TRPA1 channel in sensory neurons \u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e. Furthermore, TRPA1 stimulation enhanced glutamate release onto the nucleus tractus solitarii in the literature\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e, which points once more to the direction of the involvement of glutamate related mechanisms.\u003c/p\u003e \u003cp\u003eAltogether, our data underscores the importance of amino acid and carbohydrate metabolism for immediate energy needs, alongside lipid and redox pathways for sustained adaptation. This dynamic prioritization of metabolic pathways reflects the body's adaptability to acute stress\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e,\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. Beyond the importance of understanding such molecular processes triggered to cope with pain and stress, our neural network predictions (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) suggest that such rapid shifts in exhaled metabolic patterns could potentially be harnessed in a clinical context to gauge pain more objectively than current practice.\u003c/p\u003e \u003cp\u003eIt is important to acknowledge and consider several limitations associated with this study. Firstly, the inclusion of a relatively small number of sites and subjects compared to the generated abundance of datapoints in -omics fields, may have led to underpowered statistical analyses. Secondly, the identification of metabolites in breath samples relied on database matching using accurate mass measurements within a narrow range of ± 1 ppm. While this approach provides a preliminary identification of metabolites, it lacks the comprehensive characterization that can be achieved through mass separation techniques. Consequently, caution should be exercised when interpreting the biochemical significance of the identified metabolites. Further validation using additional analytical methods, such as tandem mass spectrometry or nuclear magnetic resonance spectroscopy, would enhance the confidence and accuracy of the metabolite identifications. Thirdly, pain manifests in various forms, and without additional studies, it remains uncertain whether our findings are exclusively attributable to pain. Other factors, such as cold exposure, stress, fluctuations in blood pressure, and related physiological responses, may also contribute. However, this overlap also serves as a means of defining pain within a broader physiological and contextual framework. Fourthly, the specific involvement of metabolic pathways and metabolites in receptor activators and pain processing is grounded in sparse and possibly underpowered studies and their precise roles in pain modulation, nociception, and noxious cold sensation require further investigation. Fifthly, the study and understanding of metabolic pathways within pain-related contexts are still in their nascent stage. Currently, numerous potential metabolic feature hits pertaining to this subject matter are being published. However, it is yet to be determined which specific findings will emerge as definitive and significant contributions, as further investigations and scientific advancements are required to unravel the complexities of these pathways in the context of pain, and in detangling the complex nature of pain itself. Having said this, one main strength of our study in this regard lies in the fact that we included a validation cohort in an independent setting, which revealed a remarkably similar response to pain induction via CPT. The highly standardized procedures conducted in both laboratories ensured a minimization of the technical variability along with a high mass accuracy\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e,\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e\u003c/sup\u003e. Moreover, several central metabolites relevant to this study (e.g. amino acids\u003csup\u003e\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e and citric acid cycle\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e), have been previously fully characterized in exhaled breath.\u003c/p\u003e \u003cp\u003eIn summary, we show here that CPT-driven pain induces a rapid physiological response encompassing a stark metabolic shift within time scales of minutes. This was observed leveraging the potential of real-time high resolution mass spectrometry to interrogate the exhaled metabolome. The altered metabolome illustrates the complex interplay between the sympathetic nervous system, pain response, energy mobilization and blood pressure regulation. Through their various roles in neurotransmission, vascular control, and energy metabolism, they help the body adapt to stressful or painful situations. Taken together, these interlinked pathways appear to optimize energy utilization, augment neurotransmission, and balance vasoactive mediators, thus promoting both the acute pain response and the hemodynamic elevation characteristic of CPT. Finally, we confirmed that such changes in the exhaled metabolic fingerprints could be used to provide non-invasively and in real time a probability score of whether the subject was in pain. These observations were observed intially in a European cohort and further validated in a Chinese cohort, reinforcing the robustness of our observations. We hypothesize that further development of this observer-independent technique could potentially guide personalized pain management in a clinical context in the various vulnerable risk groups of patient in a chronic or acute pain setting. This method could help to avoid undertreatment in these groups of patients.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003cp\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e\n\n "},{"header":"Methods","content":"\u003ch2\u003eClinical Trial Design: Pain induction via CPT\u003c/h2\u003e\u003cp\u003ePain was experimentally induced by the CPT\u003csup\u003e\u003cspan additionalcitationids=\"CR65\" citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e–\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e. During the trial, participants underwent four exhalation measurements (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e): Two baseline sets before the CPT, one immediately after the hand withdrawal from the cold water, and a subsequent final washout measurement 15 minutes after the third measurement. Participants submerged their open and relaxed right hand and 3 cm of their wrist into a container of around 13 liters containing ice-cold water and crushed ice that was manually stirred and continuously measured for temperature of 2°C (± 1°C) using a laboratory thermometer. Participants were instructed to keep their hand and wrist submerged in the water for as long as they could tolerate the pain before removing their hand, however, a 4-minute cutoff was applied, and those who could manage to keep their hand submerged for that long were also instructed to remove it. The participants received the study information and consents by email at least 24 hours before the experimental intervention. The two information packages provided to participants described the trial in simpler, shorter, and more complex detailed versions. At the study site, the entire procedure was explained in detail, and participants’ questions were addressed before obtaining consent. Informed consent and general consent were obtained from the participants before starting measurements. Before the start of the project, ethical approvals (Swiss equivalent to institutional review boards IRBs) were registered and approved with the following endpoints.\u003c/p\u003e\u003cp\u003ePrimary endpoint: The variable of primary interest is the variation of signal intensity of mass-to-charge ratio (\u003cem\u003em/z\u003c/em\u003e) of exhaled metabolites across the conducted CPT pain analgesia-placebo effects trial.\u003c/p\u003e\u003cp\u003eSecondary endpoints:\u003c/p\u003e\u003cp\u003e \u003c/p\u003e\u003cul\u003e \u003cli\u003e \u003cp\u003eDifferentially exhaled metabolites across the pain vs. no-pain measurements.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eExhaled metabolites for differences between placebo responders and non-responders.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eExhaled metabolites association with blood pressure and heart rate.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e\u003cp\u003ePre-registrations and ethical approvals are listed in \u003cb\u003eSupplementary Table\u0026nbsp;3\u003c/b\u003e.\u003c/p\u003e\u003ch3\u003eReal-time breath metabolomics by High-Resolution Mass Spectrometry\u003c/h3\u003e\u003cp\u003eThe breath metabolomics platform consisted of a Secondary Electrospray Ionization-High Resolution Mass Spectrometry (SESI-HRMS) setup. For both study sites the same settings were chosen if not mentioned separately: The system involved a direct coupling of a high-resolution mass spectrometer (Q-Exactive Plus in the Swiss site and Q-Exactive in the Chinese site, Thermo Fisher Scientific, Germany) to an ion source (Super SESI, Fossiliontech FIT, Spain).\u003c/p\u003e\u003cp\u003eA 20 µm capillary emitter (Fossiliontech FIT, Spain) was employed to generate an electrospray of water (0.1% v/v formic acid; LiChrosolv, hypergrade for LC-MS, 1.59013.2500, Millipore). The pressure within the electrospray reservoir was maintained at a pressure of 1.3 bar for the Swiss site and 0.8bar for the Chinese site respectively. Electrospray voltages were set at 2.8 kV in positive and negative mode.\u003c/p\u003e\u003cp\u003eThe ionization chamber and sample line were set to temperatures of 90°C and 130°C, respectively. The Orbitrap capillary temperature was set to 275°C, the sheath gas flow rate was set at 60 AU, and the S-lens RF level was set to 55.0 V. To capture mass spectra, the Thermo Exactive Plus Tune software (5.0.0.38 for the Swiss site and 2.8.1.2806 for the Chinese site) was utilized in full scan mode with a resolving power of 140,000 at m/z 200. The scan parameters included a range of 70-1000 m/z, selection of positive or negative polarity, 2 microscans, an ACG target of 1e6, and a maximum injection duration of 500 ms.\u003c/p\u003e\u003cp\u003eThe MS was internally calibrated by enabling lock masses corresponding to typical background mass spectrometric contaminants\u003csup\u003e67 68\u003c/sup\u003e, and externally calibrated on a weekly schedule using a commercially available calibration solution (PierceTM Triple Quadrupole, extended mass range)\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e. Prior to any breath measurement, an external gas standard (α-terpinene at 100 ppb | Dalian Special Gases Co. Ltd, Dalian, China)\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e was used to test the instrument's sensitivity. The so-called Nelson criteria in process control were employed to validate if the instrument performance was within the “in-control” range using historical data on the same gas standard\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e. Once this suitability test was passed, the subjects were invited to perform the breath test. The method followed the standardized approach described previously\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e,\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e. In short, an exhalation set includes six exhalations in positive and six exhalations in negative mode. The exhalations were controlled by a mass flow controller and guided by CO\u003csub\u003e2\u003c/sub\u003e exhalation. The process of one set took approximately seven minutes.\u003c/p\u003e\u003ch2\u003eData analysis\u003c/h2\u003e\u003cp\u003ePre-processing of mass spectrometric data\u003c/p\u003e\u003cp\u003eThe preprocessing of file spectra was undertaken as per our patented pipeline\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e: The acquired *.RAW file spectra were preprocessed using the in-house developed SESI-HRMS_Analysis_Toolbox (v 5.2.0) in MATLAB (version R2023a,b-2024a,b; MathWorks Inc., USA) and C# environment. In summary, exhalation time windows including only the end-tidal fraction were determined by the intervals when CO\u003csub\u003e2\u003c/sub\u003e concentrations were greater than 3% (as measured by Exhalion capnograph). Average raw centroid and profile mass spectra during such exhalation scans were computed using in-house C# console apps based on Thermo Fisher Scientific’s RawFileReader (version 5.0.0.38). Centroid and profile mass spectra were recalibrated to achieve mass accuracies within ± 1 ppm. Subsequently, apodization of artifact satellite peaks\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e was accomplished. Finally, the centroid data set was binned within ± 1 ppm using MATLAB’s \u003cem\u003eksdensity\u003c/em\u003e function. This resulted in a final feature list of size 11,840 in positive mode and 4,446 in negative mode.\u003c/p\u003e\u003cp\u003ePost-processing of mass spectrometric data\u003c/p\u003e\u003cp\u003eFirstly, we reduced the sparsity of the data matrix by dropping all mass spectral features with at least 50% of zero values across all samples. This reduced the number of features to 3,734 and 1,324 in positive and negative mode, respectively (i.e., 5,058 features overall). Secondly, the remaining zeros were imputed using the regression on order statistics method\u003csup\u003e\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e. To identify rapid changes in breath composition as result of CPT intervention, the two baseline and the two post-CPT measurements were fused by calculating the area under the curve (AUC) between the two pairs of measurements and normalized by the time between both measurements. Thereafter, the Log\u003csub\u003e2\u003c/sub\u003eFC between the pre- and post-CPT AUCs was computed for each feature. Log\u003csub\u003e2\u003c/sub\u003eFC were subjected to a paired t-test. Significant features defined as those for which their positive false discovery rate\u003csup\u003e\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e (i.e., FDR) ≤ 0.01 and the |Log\u003csub\u003e2\u003c/sub\u003eFC| ≥ 1.5.\u003c/p\u003e\u003ch2\u003eBiological interpretation: database query, correlation, and enrichment analysis\u003c/h2\u003e\u003cp\u003eWe matched the final features to molecular formulas with adducts M + H [1+], M(C13) + H [1+], M-H [1-], and M(C13)-H [1-] and a ppm error window of +-1, using a list that contains the “Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry all possible molecular formulas“\u003csup\u003e\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e. Further, we utilized the Human Metabolome Database (HMDB) data set ‘serum metabolites (5.0)’\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e to map our molecular formulas onto HMDB accession, taxonomy class, pathways, and metabolite descriptors.\u003c/p\u003e\u003cp\u003eCorrelation networks were computed using Pearson correlation and visualized using Cytoscape\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u003c/sup\u003e upon thresholding to rho \u0026gt; 0.7. Using this criterion, we proceeded to identify pairs of nodes that satisfied the threshold condition in both study sites. Only those node pairs whose correlations exceeded the cutoff in both matrices were selected for further analysis. The resulting pairs of nodes, fulfilling the aforementioned condition, were then organized into a tabular format representing the nodes in interaction that exhibit strong correlations across both discovery and validation study sites.\u003c/p\u003e\u003cp\u003ePathway enrichment analysis was performed using mummichog algorithm implemented in MetaboAnalystR (version 4.0.0) using hsa_kegg database and abovementioned adduct and tolerances. In cases of overlapping peaks, we assigned a lower probability weight of 0.001 to the corresponding entries in the vector dummy Ps which was otherwise initialized to 1 for all data points, ensuring differentiation between unique and overlapping signals.\u003c/p\u003e\u003cp\u003ePain prediction using neural networks\u003c/p\u003e\u003cp\u003eThe prediction of pre-CPT vs post-CPT (i.e., no pain vs. pain) breath mass spectra was accomplished using a multi-layer perceptron network in a Python environment. Firstly, the profile mass spectra were linearly interpolated, and the positive and negative ion mode spectra concatenated. This resulted in a final data matrix of 160 samples (2 samples pre-CPT + 2 samples post-CPT = 4 breath samples per subject; 19 participants in discovery cohort and 21 in validation cohort). The dimensionality was reduced via principal component analysis (PCA) to 48, 64, 96 128 or 144 features. The number of input neurons of the network architecture corresponded to the available PCA features followed by two hidden layers where the sizes: [32,32], [48,48], [64,64] or [96,96] where evaluated. The data − at the participant level − was split in 90% training, 5% validation and 5% testing for cross validation. This process was repeated 50 times with random partitions in each fold and for all combinations of number of PCA features and sizes of hidden neurons. Thus, all four breath mass spectra for a given participant would be either in the training, validation or the test set but never mixed to avoid potential classification bias. The training set was subjected to PCA to reduce the dimensionality, whereby the first n scores were used as predicting features. Once the model was trained, the test set was projected onto the same PCA space of the training set and predicted using the first n scores. The pain/no-pain prediction performance was estimated by the area under the curve of the ROC curve. Upon hyperparameter optimization of the network architecture (number of PCA scores and size of the hidden layer), the best setting classification scores were obtained with 48 PCA features with 96,96 hidden layers.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePS is co-founder and board member of Deep Breath Intelligence (DBI) AG, a company that provides services in the field of breath analysis. Kapil Dev Singh is partially employed by the same company. UKBB is a shareholder of said company.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was funded by grants from Fondation Botnar (Switzerland) No. 320030_173168 (to PS) and the Swiss National Science Foundation (SNSF) PCEGP3-181300/ 501100001711-173168 (to PS). XL received funding from Guangdong Major Project of Basic and Applied Basic Research (No. 2023B0303000013), Guangdong Provincial International Science and Technology Cooperation Project (No. 2022A0505050044), and the National Natural Science Foundation of China (No. 22122603).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eBrennan, F., Carr, D. B. \u0026amp; Cousins, M. Pain Management: A Fundamental Human Right. \u003cem\u003eAnesthesia \u0026amp; Analgesia\u003c/em\u003e \u003cstrong\u003e105\u003c/strong\u003e, 205-221 (2007). https://doi.org/10.1213/01.ane.0000268145.52345.55\u003c/li\u003e\n \u003cli\u003eGaskin, D. J. \u0026amp; Richard, P. The Economic Costs of Pain in the United States. \u003cem\u003eThe Journal of Pain\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 715-724 (2012). https://doi.org/10.1016/j.jpain.2012.03.009\u003c/li\u003e\n \u003cli\u003eCordell, W. H.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e The high prevalence of pain in emergency medical care. \u003cem\u003eAm J Emerg Med\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 165-169 (2002). https://doi.org/10.1053/ajem.2002.32643\u003c/li\u003e\n \u003cli\u003eChou R, W. J., Ahmed AY, et al. Vol. Comparative Effectiveness Review, No. 240. (Agency for Healthcare Research and Quality (US), Rockville (MD), 2020).\u003c/li\u003e\n \u003cli\u003eFinley, C. R.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e What are the most common conditions in primary care? \u003cem\u003eCanadian Family Physician | Le M\u0026eacute;decin de famille canadien Systematic review\u003c/em\u003e \u003cstrong\u003e64\u003c/strong\u003e, 832-840 (2018).\u003c/li\u003e\n \u003cli\u003eChou, R.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Management of Postoperative Pain: A Clinical Practice Guideline From the American Pain Society, the American Society of Regional Anesthesia and Pain Medicine, and the American Society of Anesthesiologists\u0026apos; Committee on Regional Anesthesia, Executive Committee, and Administrative Council. \u003cem\u003eThe Journal of Pain\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 131-157 (2016). https://doi.org/https://doi.org/10.1016/j.jpain.2015.12.008\u003c/li\u003e\n \u003cli\u003eSchofield, P.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Evidence-based clinical practice guidelines on the management of pain in older people - a summary report. \u003cem\u003eBr J Pain\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 6-13 (2022). https://doi.org/10.1177/2049463720976155\u003c/li\u003e\n \u003cli\u003eBeltramini, A., Milojevic, K. \u0026amp; Pateron, D. Pain Assessment in Newborns, Infants, and Children. \u003cem\u003ePediatr Ann\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, e387-e395 (2017). https://doi.org/10.3928/19382359-20170921-03\u003c/li\u003e\n \u003cli\u003eCatananti, C. \u0026amp; Gambassi, G. Pain assessment in the elderly. \u003cem\u003eSurg Oncol\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 140-148 (2010). https://doi.org/10.1016/j.suronc.2009.11.010\u003c/li\u003e\n \u003cli\u003eCook, A. K., Niven, C. A. \u0026amp; Downs, M. G. Assessing the pain of people with cognitive impairment. \u003cem\u003eInt J Geriatr Psychiatry\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 421-425 (1999). https://doi.org/10.1002/(sici)1099-1166(199906)14:6\u0026lt;421::aid-gps932\u0026gt;3.0.co;2-y\u003c/li\u003e\n \u003cli\u003eAlderson, S. M. \u0026amp; McKechnie, S. R. Unrecognised, undertreated, pain in ICU\u0026mdash;Causes, effects, and how to do better. \u003cem\u003eOpen Journal of Nursing\u003c/em\u003e \u003cstrong\u003e03\u003c/strong\u003e, 108-113 (2013). https://doi.org/10.4236/ojn.2013.31014\u003c/li\u003e\n \u003cli\u003eHullett, B.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Monitoring electrical skin conductance: a tool for the assessment of postoperative pain in children? \u003cem\u003eAnesthesiology\u003c/em\u003e \u003cstrong\u003e111\u003c/strong\u003e, 513-517 (2009). https://doi.org/10.1097/ALN.0b013e3181b27c18\u003c/li\u003e\n \u003cli\u003eDalal, P. G.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Analysis of acute pain scores and skin conductance measurements in infants. \u003cem\u003eEarly Human Development\u003c/em\u003e \u003cstrong\u003e89\u003c/strong\u003e, 153-158 (2013). https://doi.org/https://doi.org/10.1016/j.earlhumdev.2012.09.008\u003c/li\u003e\n \u003cli\u003eHu, J., Harrold, J., Squires, J. E., Modanloo, S. \u0026amp; Harrison, D. The validity of skin conductance for assessing acute pain in mechanically ventilated infants: A cross‐sectional observational study. \u003cem\u003eEuropean Journal of Pain\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1994-2006 (2021). https://doi.org/10.1002/ejp.1816\u003c/li\u003e\n \u003cli\u003eBrown, D. N. Pain assessment in the recovery room. \u003cem\u003eJ Perioper Pract\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 480-489 (2008). https://doi.org/10.1177/175045890801801103\u003c/li\u003e\n \u003cli\u003eBreivik, H.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Assessment of pain. \u003cem\u003eBritish Journal of Anaesthesia\u003c/em\u003e \u003cstrong\u003e101\u003c/strong\u003e, 17-24 (2008). https://doi.org/10.1093/bja/aen103\u003c/li\u003e\n \u003cli\u003eAldington, D. \u0026amp; Eccleston, C. Evidence-Based Pain Management: Building on the Foundations of Cochrane Systematic Reviews. \u003cem\u003eAmerican Journal of Public Health\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 46-49 (2019). https://doi.org/10.2105/ajph.2018.304745\u003c/li\u003e\n \u003cli\u003eLink, H., Fuhrer, T., Gerosa, L., Zamboni, N. \u0026amp; Sauer, U. Real-time metabolome profiling of the metabolic switch between starvation and growth. \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 1091-1097 (2015). https://doi.org/10.1038/nmeth.3584\u003c/li\u003e\n \u003cli\u003eRohde, P. D., Kristensen, T. N., Sarup, P., Munoz, J. \u0026amp; Malmendal, A. Prediction of complex phenotypes using the Drosophila melanogaster metabolome. \u003cem\u003eHeredity (Edinb)\u003c/em\u003e \u003cstrong\u003e126\u003c/strong\u003e, 717-732 (2021). https://doi.org/10.1038/s41437-021-00404-1\u003c/li\u003e\n \u003cli\u003eZhou, S.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Systems genetics of the Drosophila metabolome. \u003cem\u003eGenome Res\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 392-405 (2020). https://doi.org/10.1101/gr.243030.118\u003c/li\u003e\n \u003cli\u003eCastelli, F. A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Metabolomics for personalized medicine: the input of analytical chemistry from biomarker discovery to point-of-care tests. \u003cem\u003eAnal Bioanal Chem\u003c/em\u003e \u003cstrong\u003e414\u003c/strong\u003e, 759-789 (2022). https://doi.org/10.1007/s00216-021-03586-z\u003c/li\u003e\n \u003cli\u003eJacob, M., Lopata, A. L., Dasouki, M. \u0026amp; Abdel Rahman, A. M. Metabolomics toward personalized medicine. \u003cem\u003eMass Spectrom Rev\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 221-238 (2019). https://doi.org/10.1002/mas.21548\u003c/li\u003e\n \u003cli\u003eKogelman, L. J. A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Multi-omics to predict changes during cold pressor test. \u003cem\u003eBMC Genomics\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 759 (2022). https://doi.org/10.1186/s12864-022-08981-z\u003c/li\u003e\n \u003cli\u003eHsu, W. H.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Metabolomic and proteomic characterization of sng and pain phenotypes in fibromyalgia. \u003cem\u003eEur J Pain\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 445-462 (2022). https://doi.org/10.1002/ejp.1871\u003c/li\u003e\n \u003cli\u003eL\u0026ouml;tsch, J.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Smell of pain: intersection of nociception and olfaction. \u003cem\u003ePAIN\u003c/em\u003e \u003cstrong\u003e157\u003c/strong\u003e, 2152-2157 (2016). https://doi.org/10.1097/j.pain.0000000000000599\u003c/li\u003e\n \u003cli\u003eModir, J. G. \u0026amp; Wallace, M. S. in \u003cem\u003eAnalgesia: Methods and Protocols\u003c/em\u003e (ed Arpad Szallasi) 165-168 (Humana Press, 2010).\u003c/li\u003e\n \u003cli\u003eLamotte, G., Boes, C. J., Low, P. A. \u0026amp; Coon, E. A. The expanding role of the cold pressor test: a brief history. \u003cem\u003eClinical Autonomic Research\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 153-155 (2021). https://doi.org/10.1007/s10286-021-00796-4\u003c/li\u003e\n \u003cli\u003eFanninger, S., Plener, P. L., Fischer, M. J. M., Kothgassner, O. D. \u0026amp; Goreis, A. Water temperature during the cold pressor test: A scoping review. \u003cem\u003ePhysiology \u0026amp; Behavior\u003c/em\u003e \u003cstrong\u003e271\u003c/strong\u003e, 114354 (2023). https://doi.org/https://doi.org/10.1016/j.physbeh.2023.114354\u003c/li\u003e\n \u003cli\u003eWishart, D. S.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e HMDB 5.0: the Human Metabolome Database for 2022. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, D622-d631 (2022). https://doi.org/10.1093/nar/gkab1062\u003c/li\u003e\n \u003cli\u003eWesterhuis, J. A., Van Velzen, E. J. J., Hoefsloot, H. C. J. \u0026amp; Smilde, A. K. Discriminant Q2 (DQ2) for improved discrimination in PLSDA models. \u003cem\u003eMetabolomics\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 293-296 (2008). https://doi.org/10.1007/s11306-008-0126-2\u003c/li\u003e\n \u003cli\u003eHerdin, M., Czink, N., Ozcelik, H. \u0026amp; Bonek, E. in \u003cem\u003e2005 IEEE 61st Vehicular Technology Conference.\u003c/em\u003e 136-140 Vol. 131.\u003c/li\u003e\n \u003cli\u003eBerg, J. M., Tymoczko, J. L., Gatto, J. G. J. \u0026amp; Stryer, L. \u003cem\u003eBiochemistry\u003c/em\u003e. Eighth edition, fourth printing edn, (Macmillan [etc.], 2015).\u003c/li\u003e\n \u003cli\u003eKanehisa, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Data, information, knowledge and principle: back to metabolism in KEGG. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, D199-205 (2014). https://doi.org/10.1093/nar/gkt1076\u003c/li\u003e\n \u003cli\u003eLi, S.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Predicting network activity from high throughput metabolomics. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e1003123 (2013). https://doi.org/10.1371/journal.pcbi.1003123\u003c/li\u003e\n \u003cli\u003eSolana, M. J.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Assessment of pain in critically ill children. Is cutaneous conductance a reliable tool? \u003cem\u003eJournal of Critical Care\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 481-485 (2015). https://doi.org/https://doi.org/10.1016/j.jcrc.2015.01.008\u003c/li\u003e\n \u003cli\u003eRaja, S. N.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e The revised International Association for the Study of Pain definition of pain: concepts, challenges, and compromises. \u003cem\u003ePAIN\u003c/em\u003e \u003cstrong\u003e161\u003c/strong\u003e, 1976-1982 (2020). https://doi.org/10.1097/j.pain.0000000000001939\u003c/li\u003e\n \u003cli\u003ePark, S. G., Ewalt, K. L. \u0026amp; Kim, S. Functional expansion of aminoacyl-tRNA synthetases and their interacting factors: new perspectives on housekeepers. \u003cem\u003eTrends in Biochemical Sciences\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 569-574 (2005). https://doi.org/10.1016/j.tibs.2005.08.004\u003c/li\u003e\n \u003cli\u003eEspinosa-Diez, C.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Antioxidant responses and cellular adjustments to oxidative stress. \u003cem\u003eRedox Biol.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 183-197 (2015). https://doi.org/https://doi.org/10.1016/j.redox.2015.07.008\u003c/li\u003e\n \u003cli\u003eDalile, B., Van Oudenhove, L., Vervliet, B. \u0026amp; Verbeke, K. The role of short-chain fatty acids in microbiota-gut-brain communication. \u003cem\u003eNat. Rev. Gastroenterol. Hepatol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 461-478 (2019). https://doi.org/10.1038/s41575-019-0157-3\u003c/li\u003e\n \u003cli\u003eLatremoliere, A. \u0026amp; Woolf, C. J. Central Sensitization: A Generator of Pain Hypersensitivity by Central Neural Plasticity. \u003cem\u003eThe Journal of Pain\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 895-926 (2009). https://doi.org/10.1016/j.jpain.2009.06.012\u003c/li\u003e\n \u003cli\u003ePalmer, R. M. J., Ashton, D. S. \u0026amp; Moncada, S. Vascular endothelial cells synthesize nitric oxide from L-arginine. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e333\u003c/strong\u003e, 664-666 (1988). https://doi.org/10.1038/333664a0\u003c/li\u003e\n \u003cli\u003eClarke, H. E., Kluess, H., Adams, S., Kim, D.-H. \u0026amp; Hickner, R. Metabolomics Profiling in Response to a Cold Pressor Test. \u003cem\u003eThe FASEB Journal\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 1-1 (2020). https://doi.org/https://doi.org/10.1096/fasebj.2020.34.s1.06129\u003c/li\u003e\n \u003cli\u003eSeravalle, G. \u0026amp; Grassi, G. Sympathetic Nervous System, Hypertension, Obesity and Metabolic Syndrome. \u003cem\u003eHigh Blood Press. Cardiovasc. Prev.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 175-179 (2016). https://doi.org/10.1007/s40292-016-0137-4\u003c/li\u003e\n \u003cli\u003eHirooka, Y. Sympathetic Activation in Hypertension: Importance of the Central Nervous System. \u003cem\u003eAm. J. Hypertens.\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 914-926 (2020). https://doi.org/10.1093/ajh/hpaa074\u003c/li\u003e\n \u003cli\u003eHolton, K. F. in \u003cem\u003eGlutamate and Neuropsychiatric Disorders: Current and Emerging Treatments\u003c/em\u003e (ed Zoran M. Pavlovic) 381-402 (Springer International Publishing, 2022).\u003c/li\u003e\n \u003cli\u003ePal, M. M. Glutamate: The Master Neurotransmitter and Its Implications in Chronic Stress and Mood Disorders. \u003cem\u003eFront. Hum. Neurosci.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e (2021). https://doi.org/10.3389/fnhum.2021.722323\u003c/li\u003e\n \u003cli\u003eMazzitelli, M., Palazzo, E., Maione, S. \u0026amp; Neugebauer, V. Group II Metabotropic Glutamate Receptors: Role in Pain Mechanisms and Pain Modulation. \u003cem\u003eFront. Mol. Neurosci.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e (2018). https://doi.org/10.3389/fnmol.2018.00383\u003c/li\u003e\n \u003cli\u003eBrosnan, J. T. \u0026amp; Brosnan, M. E. Creatine: Endogenous Metabolite, Dietary, and Therapeutic Supplement. \u003cem\u003eAnnu. Rev. Nutr.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 241-261 (2007). https://doi.org/https://doi.org/10.1146/annurev.nutr.27.061406.093621\u003c/li\u003e\n \u003cli\u003eAroke, E. N. \u0026amp; Powell-Roach, K. L. The Metabolomics of Chronic Pain Conditions: A Systematic Review. \u003cem\u003eBiol Res Nurs\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 458-471 (2020). https://doi.org/10.1177/1099800420941105\u003c/li\u003e\n \u003cli\u003eTeckchandani, S., Nagana Gowda, G. A., Raftery, D. \u0026amp; Curatolo, M. Metabolomics in chronic pain research. \u003cem\u003eEur J Pain\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 313-326 (2021). https://doi.org/10.1002/ejp.1677\u003c/li\u003e\n \u003cli\u003eWoolf, C. J. \u0026amp; Thompson, S. W. N. The induction and maintenance of central sensitization is dependent on N-methyl-d-aspartic acid receptor activation; implications for the treatment of post-injury pain hypersensitivity states. \u003cem\u003ePain\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 293-299 (1991). https://doi.org/https://doi.org/10.1016/0304-3959(91)90100-C\u003c/li\u003e\n \u003cli\u003eJang, J. H., Kim, D. W., Sang Nam, T., Se Paik, K. \u0026amp; Leem, J. W. Peripheral glutamate receptors contribute to mechanical hyperalgesia in a neuropathic pain model of the rat. \u003cem\u003eNeuroscience\u003c/em\u003e \u003cstrong\u003e128\u003c/strong\u003e, 169-176 (2004). https://doi.org/https://doi.org/10.1016/j.neuroscience.2004.06.040\u003c/li\u003e\n \u003cli\u003eCairns, B. E., Hu, J. W., Arendt-Nielsen, L., Sessle, B. J. \u0026amp; Svensson, P. Sex-Related Differences in Human Pain and Rat Afferent Discharge Evoked by Injection of Glutamate Into the Masseter Muscle. \u003cem\u003eJournal of Neurophysiology\u003c/em\u003e \u003cstrong\u003e86\u003c/strong\u003e, 782-791 (2001). https://doi.org/10.1152/jn.2001.86.2.782\u003c/li\u003e\n \u003cli\u003eMacDonald, D. I., Wood, J. N. \u0026amp; Emery, E. C. Molecular mechanisms of cold pain. \u003cem\u003eNeurobiology of Pain\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 100044 (2020). https://doi.org/https://doi.org/10.1016/j.ynpai.2020.100044\u003c/li\u003e\n \u003cli\u003eStory, G. M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e ANKTM1, a TRP-like channel expressed in nociceptive neurons, is activated by cold temperatures. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e112\u003c/strong\u003e, 819-829 (2003). https://doi.org/10.1016/s0092-8674(03)00158-2\u003c/li\u003e\n \u003cli\u003eStory, G. M. \u0026amp; Gereau, R. W. Numbing the senses: role of TRPA1 in mechanical and cold sensation. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, 177-180 (2006). https://doi.org/10.1016/j.neuron.2006.04.009\u003c/li\u003e\n \u003cli\u003eSun, B., Bang, S.-I. \u0026amp; Jin, Y.-H. Transient receptor potential A1 increase glutamate release on brain stem neurons. \u003cem\u003eNeuroReport\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 1002-1006 (2009). https://doi.org/10.1097/WNR.0b013e32832d2219\u003c/li\u003e\n \u003cli\u003eSprague, J. E. \u0026amp; Arbel\u0026aacute;ez, A. M. Glucose counterregulatory responses to hypoglycemia. \u003cem\u003ePediatr Endocrinol Rev\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 463-473; quiz 474-465 (2011).\u003c/li\u003e\n \u003cli\u003eSapolsky, R. M. Stress and plasticity in the limbic system. \u003cem\u003eNeurochem Res\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1735-1742 (2003). https://doi.org/10.1023/a:1026021307833\u003c/li\u003e\n \u003cli\u003eSingh, K. D.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Standardization procedures for real-time breath analysis by secondary electrospray ionization high-resolution mass spectrometry. \u003cem\u003eAnal Bioanal Chem\u003c/em\u003e \u003cstrong\u003e411\u003c/strong\u003e, 4883-4898 (2019). https://doi.org/10.1007/s00216-019-01764-8\u003c/li\u003e\n \u003cli\u003eGisler, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e An Interoperability Framework for Multicentric Breath Metabolomic Studies. \u003cem\u003eiScience\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 105557 (2022). https://doi.org/https://dx.doi.org/10.2139/ssrn.4076338\u003c/li\u003e\n \u003cli\u003eGarc\u0026iacute;a-G\u0026oacute;mez, D.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Real-Time Quantification of Amino Acids in the Exhalome by Secondary Electrospray Ionization\u0026ndash;Mass Spectrometry: A Proof-of-Principle Study. \u003cem\u003eClin. Chem.\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 1230-1237 (2016). https://doi.org/10.1373/clinchem.2016.256909\u003c/li\u003e\n \u003cli\u003eTejero Rioseras, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Real-Time Monitoring of Tricarboxylic Acid Metabolites in Exhaled Breath. \u003cem\u003eAnal. Chem.\u003c/em\u003e \u003cstrong\u003e90\u003c/strong\u003e, 6453-6460 (2018). https://doi.org/10.1021/acs.analchem.7b04600\u003c/li\u003e\n \u003cli\u003eFontanillas, P., Kless, A., Team, a. R., Bothmer, J. \u0026amp; Tung, J. Y. Genome-wide association study of pain sensitivity assessed by questionnaire and the cold pressor test. \u003cem\u003ePAIN\u003c/em\u003e \u003cstrong\u003e163\u003c/strong\u003e, 1763-1776 (2022). https://doi.org/10.1097/j.pain.0000000000002568\u003c/li\u003e\n \u003cli\u003eMacLachlan, C., Shipton, E. A. \u0026amp; Wells, J. E. The Cold Pressor Test as a Predictor of Prolonged Postoperative Pain, a Prospective Cohort Study. \u003cem\u003ePain and Therapy\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 203-213 (2016). https://doi.org/10.1007/s40122-016-0056-z\u003c/li\u003e\n \u003cli\u003eHuang, M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Early sympathetic neural responses during a cold pressor test linked to pain perception. \u003cem\u003eClinical Autonomic Research\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 215-224 (2021). https://doi.org/10.1007/s10286-019-00635-7\u003c/li\u003e\n \u003cli\u003eKeller, B. O., Sui, J., Young, A. B. \u0026amp; Whittal, R. M. Interferences and contaminants encountered in modern mass spectrometry. \u003cem\u003eAnalytica Chimica Acta\u003c/em\u003e \u003cstrong\u003e627\u003c/strong\u003e, 71-81 (2008). https://doi.org/https://doi.org/10.1016/j.aca.2008.04.043\u003c/li\u003e\n \u003cli\u003eSchlosser, A. \u0026amp; Volkmer-Engert, R. Volatile polydimethylcyclosiloxanes in the ambient laboratory air identified as source of extreme background signals in nanoelectrospray mass spectrometry. \u003cem\u003eJournal of Mass Spectrometry\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 523-525 (2003). https://doi.org/https://doi.org/10.1002/jms.465\u003c/li\u003e\n \u003cli\u003eZeng, J., Christen, A., Dev Singh, K., Frey, U. \u0026amp; Sinues, P. Comparison of Plasma Ionization- and Secondary Electrospray Ionization- High-resolution Mass Spectrometry for Real-time Breath Analysis. \u003cem\u003eCHIMIA\u003c/em\u003e \u003cstrong\u003e76\u003c/strong\u003e, 127 (2022). https://doi.org/10.2533/chimia.2022.127\u003c/li\u003e\n \u003cli\u003eLiu, C.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Quantification of volatile organic compounds by secondary electrospray ionization-high resolution mass spectrometry. \u003cem\u003eAnalytica Chimica Acta\u003c/em\u003e \u003cstrong\u003e1180\u003c/strong\u003e, 338876 (2021). https://doi.org/https://doi.org/10.1016/j.aca.2021.338876\u003c/li\u003e\n \u003cli\u003eNelson, L. S. The Shewhart Control Chart\u0026mdash;Tests for Special Causes. \u003cem\u003eJournal of Quality Technology\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 237-239 (1984). https://doi.org/10.1080/00224065.1984.11978921\u003c/li\u003e\n \u003cli\u003eGisler, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e An interoperability framework for multicentric breath metabolomic studies. \u003cem\u003eiScience\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 105557 (2022). https://doi.org/10.1016/j.isci.2022.105557\u003c/li\u003e\n \u003cli\u003eSingh, K. D. G., Amanda; Osswald, Martin; Kohler, Malcolm; Van Den Anker, Johannes; Von Der Weid, Nicolas; Frey, Urs; Datta, Alexandre; Sinues, Pablo. European Patent EP4181781 - BREATH-BASED THERAPEUTIC DRUG MONITORING METHOD. (2021).\u003c/li\u003e\n \u003cli\u003eSingh, K. D. S., Pablo. European Patent EP4370025 - METHOD FOR PROCESSING MASS SPECTROMETRY DATA OBTAINED FROM BREATH GAS. (2021).\u003c/li\u003e\n \u003cli\u003eMitchell, J. M.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e New methods to identify high peak density artifacts in Fourier transform mass spectra and to mitigate their effects on high-throughput metabolomic data analysis. \u003cem\u003eMetabolomics\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 125 (2018). https://doi.org/10.1007/s11306-018-1426-9\u003c/li\u003e\n \u003cli\u003eGelman, A., Hill, J. \u0026amp; Vehtari, A. \u003cem\u003eRegression and Other Stories\u003c/em\u003e. (Cambridge University Press, 2020).\u003c/li\u003e\n \u003cli\u003eStorey, J. D. A direct approach to false discovery rates. \u003cem\u003eJ. Roy. Stat. Soc. Ser. B. (Stat. Method.)\u003c/em\u003e \u003cstrong\u003e64\u003c/strong\u003e, 479-498 (2002). https://doi.org/10.1111/1467-9868.00346\u003c/li\u003e\n \u003cli\u003eKind, T. \u0026amp; Fiehn, O. Seven Golden Rules for heuristic filtering of molecular formulas obtained by accurate mass spectrometry. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 105 (2007). https://doi.org/10.1186/1471-2105-8-105\u003c/li\u003e\n \u003cli\u003eGustavsen, J. A., Pai, S., Isserlin, R., Demchak, B. \u0026amp; Pico, A. R. RCy3: Network biology using Cytoscape from within R. \u003cem\u003eF1000Res\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1774 (2019). https://doi.org/10.12688/f1000research.20887.3\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6048423/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6048423/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe objectification of pain presents a significant clinical challenge, particularly in children, elderly individuals, patients with disabilities and unconscious patients. It is critically important to accurately assess pain in these populations due to the heightened risk of undertreatment. Using the cold pressor test (CPT) as a pain induction model, we combined real-time breath metabolomics with pathway analysis to uncover metabolic shifts. Exhaled breath was analyzed in a discovery cohort (n=19) and validated in an independent cohort (n=21) using secondary electrospray ionization-high-resolution mass spectrometry (SESI-HRMS). Within 15 minutes of CPT, over 400 conserved mass spectral features were significantly altered across both cohorts. Pathway analysis highlighted shifts in aminoacyl-tRNA biosynthesis, cysteine/methionine metabolism, butanoate metabolism, and arginine/proline metabolism. Arginine and glutamate, key contributors to nitric oxide production and nociceptive signaling, exhibited consistent upregulation. Neural network classifiers achieved robust differentiation between pre- and post-CPT profiles (AUC=0.856), showcasing breath metabolomics as a promising observer independent, and objective tool for real-time pain assessment. To validate universal mechanistic relevance of the findings, we compared them to findings of chronic pain studies revealing consistencies in amino acid and neurotransmitter-related pathways. This study provides novel insights into the metabolic basis of acute pain and positions breath metabolomics as a viable approach for dynamic, observer independent monitoring pain in vulnerable patient groups. Future research must determine if these new insights into mechanistic pathways can inform patient- and disease-specific pain management strategies.\u003c/p\u003e","manuscriptTitle":"Pain induces a rapid characteristic metabolic signature detectable in breath","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-04 08:54:16","doi":"10.21203/rs.3.rs-6048423/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4e44baaa-0e9d-400d-8275-b38dd291c24a","owner":[],"postedDate":"March 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":45002339,"name":"Health sciences/Biomarkers/Diagnostic markers"},{"id":45002340,"name":"Health sciences/Medical research/Translational research"}],"tags":[],"updatedAt":"2026-05-07T02:02:27+00:00","versionOfRecord":{"articleIdentity":"rs-6048423","link":"https://doi.org/10.1016/j.isci.2026.115857","journal":{"identity":"iscience","isVorOnly":true,"title":"iScience"},"publishedOn":"2026-04-22 00:00:00","publishedOnDateReadable":"April 22nd, 2026"},"versionCreatedAt":"2025-03-04 08:54:16","video":"","vorDoi":"10.1016/j.isci.2026.115857","vorDoiUrl":"https://doi.org/10.1016/j.isci.2026.115857","workflowStages":[]},"version":"v1","identity":"rs-6048423","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6048423","identity":"rs-6048423","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}