All-Flexible Chronoepifluidic Nanoplasmonic Patch Allows Label-free Sweat Profiling

All-Flexible Chronoepifluidic Nanoplasmonic Patch Allows Label-free Sweat Profiling | 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 All-Flexible Chronoepifluidic Nanoplasmonic Patch Allows Label-free Sweat Profiling Ki-Hun Jeong, Jaehun Jeon, Sangyeon Lee, Seongok Chae, Joo Hoon Lee, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5624954/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 27 Aug, 2025 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Wearable sweat sensors allow non-invasive metabolic profiling for timely intervention in proactive healthcare. However, molecular recognition elements in conventional sensors still hinder a comprehensive understanding of an individual's unique physiological profile. Here we report an all-flexible chronoepifluidic surface-enhanced Raman spectroscopy (SERS) patch (CEP-SERS patch) for label-free sweat profiling. The CEP-SERS patch features the integration of nanoplasmonics and functional microfluidics for precise chronological profiling of metabolites. An ultrathin fluorocarbon film facilitates large-area nanofabrication of plasmonic structures on a functional microfluidic channel via low-temperature solid-state dewetting of a thin silver film. The CEP-SERS patch facilitates conformal contact on human skin and SERS detection of diverse metabolites from sequentially sampled sweat. Machine-learned quantification of metabolites including lactate, uric acid, and tyrosine has successfully profiled SERS detection of sweat during assorted physical activities. This CEP-SERS patch can provide a new strategy for delineating the physiological phenotype of individuals in personalized healthcare. Biological sciences/Biotechnology/Nanobiotechnology/Nanofabrication and nanopatterning Physical sciences/Nanoscience and technology/Nanobiotechnology/Microfluidics Physical sciences/Engineering/Biomedical engineering Physical sciences/Materials science/Materials for optics/Nanophotonics and plasmonics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Metabolic phenotyping is pivotal in precision medicine, revealing individual health traits for personalized interventions [1]. Metabotypes surpass genetic predispositions, offering valuable insights into the diversity of individuals influenced by extrinsic determinants such as behavioral patterns, regimens, or gut microbial activity [2,3]. The interrelation of the metabolite reflects precision health status through metabolic pathway alterations induced by physiological changes or disorders. [4–8]. In particular, transient metabolic alterations provide a physiological cue for proactive healthcare. Such cues facilitate the instant recognition of physiological disorders to promote optimal health outcomes. For instance, postprandial metabolic rates provide crucial information on individual responses to dietary intake, aiding in the development of personalized nutrition plans [9]. In addition, exercise or lifecycle routines can be tailored by monitoring daily activity-induced metabolite fluctuations [10]. However, conventional approaches profile static metabolite in blood [11,12] or urine [13–15], which poses challenges in dynamic metabolic profiling. Recent biosensing wearables profile the metabolic kinetics in biological fluids such as tear fluid [16,17], saliva [18], or sweat [19,20]. Sweat, unlike others, exhibits chemical abundance [21,22] and simple sample collection with less contamination [23], thus allowing in-situ profiling of metabolite alteration in proactive healthcare [24]. For example, analyzing lactate dynamics suggests optimal exercise routines by evaluating the lactate threshold or maximal lactate steady state during physical activities [25–27]. In addition, postprandial changes of branched-chain amino acid or uric acid provide prognostic cues for metabolic syndrome [28–30] or gout [31], respectively. Moreover, cortisol levels [32–34] indicate psychological stress responses during daily life. Furthermore, sweat is easily collected by epidermal interfaces with simple absorbents [35–37], or microfluidic patches [38–40], exhibiting high compatibility with assorted wearable sensors. Such epidermal sweat sensors often utilize electrochemical [26,27,30,31,33,34,36,38,41] or colorimetric approaches [42–45]. However, molecular recognition elements such as antibodies or enzymes on active sensing sites still hinder multiple and unveiled biomarkers detection for a comprehensive understanding of an individual's unique physiological profile. Surface-enhanced Raman spectroscopy (SERS) allows label-free and quantitative detections of diverse biochemicals [46–48]. Plasmonic nanostructures are integrated into flexible epidermal patches for label-free metabolic profiling through sweat. [48,49,50,51] Furthermore, recent advancements in plasmonic patches combine with epidermal microfluidic channels for not only transient analysis [49–57] but also in-situ collection and chronological profiling of metabolites [58,59,60]. For instance, rigid SERS substrates are frequently integrated into elastomeric microfluidic patches [58,61] to analyze metabolic alterations [58]. Partial rigidity of the SERS substrate still constrains conformal skin contact and mechanical durability in wearable applications, whereas plasmonic paper-based microfluidic sensors feature high flexibility thus overcoming such technical limitations. However, such sensors still lack delicate microfluidic control [48,52,55,62,63], sensitivity, and specificity [59,64], thus restricting the direct quantification of multiple analytes in complex sweat. In particular, such SERS patches encounter compatibility issues with sophisticated functional microfluidics such as capillary bursting valves [65–67], facilitating sequential sampling for sweat profiling. As a result, the all-flexible epifluidic SERS patch still remains a technical challenge in precise and multiplexed sweat profiling. Here we report an all-flexible chronoepifluidic SERS patch (CEP-SERS patch) for label-free profiling of sweat metabolites. The CEP-SERS patch consists of a plasmofluidic channel layer (PCL) for sweat collection, storage, and SERS analysis, and a dermal contact layer (DCL) for skin attachment (Fig. 1 a). PCL includes a flexible plasmonic SERS substrate and a microfluidic sequential sampler. The SERS substrate features plasmonic nanoislands on ultrathin fluorocarbon-coated PDMS membrane, driven by low-temperature solid-state dewetting of thin silver film. The ultrathin fluorocarbon film effectively dewets the thin metal film on the sequential sampler, resulting in plasmonic structures with strong electromagnetic hotspots for highly sensitive SERS analysis. In addition, the sequential sampler serially collects the sweat via the microfluidic channel with capillary bursting valves, which spatially separate the sweat over time. Furthermore, DCL contains medical adhesive with a sweat collection port and interconnects the PCL on the skin for stable sweat collection. Finally, the CEP-SERS patch allows machine-learned label-free quantification of multiple metabolites in chrono-sampled sweat, thus quantitatively profiling sweat over time during physical activities. Such all-flexible feature ensures conformal contact for on-skin chronological sweat collection and label-free quantification of metabolites. The CEP-SERS patch was fabricated by integrating micro- and nanofabricated PCL and DCL (Fig. 1 b). First, an ultrathin fluorocarbon layer was coated on the polydimethylsiloxane (PDMS) chronoepifluidic sweat sampler via atmospheric pressure chemical vapor deposition (AP-CVD) (S1). The fluorocarbon coating significantly reduces the surface energy and increases the surface roughness, thereby facilitating the low-temperature solid-state dewetting of an Ag thin film. Subsequently, a 10 nm-thick Ag thin film was thermally evaporated onto the fluorocarbon-coated sweat sampler (S2). The evaporated Ag thin film was thermally dewetted at 160°C for 30 minutes to form Ag nanoislands (S3). Steps S2 and S3 were repeated to enhance the plasmonic hotspots on the PCL. Finally, an air-venting outlet was punched on the PCL attached to the DCL with a sweat port for encapsulation and skin attachment (S4). Figure 1 c and 1 d present optical images of the fabricated PCL, including an inset scanning electron microscopy (SEM) image of the Ag nanoislands, and the CEP-SERS patch on the skin. Details and stepwise optical images of the fabrication procedure are described in the Supplementary Materials (Fig. S1 ). Result All-flexible SERS Substrate: Ag nanoislands on ultrathin fluorocarbon The plasmonic properties of Ag nanoislands on the CEP-SERS patch are precisely controlled by the thicknesses of fluorocarbon and Ag thin film. Figure 2 a presents SEM images (left) and E-field distribution (middle) of Ag nanoislands, calculated by the finite-difference time-domain method (Lumerical FDTD Solutions) method. The structural features (right top) and absorption properties (right bottom) are shown for 10 nm Ag film on PDMS without fluorocarbon coating (green), single-dewetted 10 nm Ag film (red), and double-dewetted 10 nm Ag film (blue) on fluorocarbon coated PDMS. Unlike bare PDMS, the fluorocarbon intermediate layer provides low surface energy, leading to an enlargement of nanoisland diameters by facilitating lower-temperature solid-state dewetting of thin film (Fig. S2). In addition, two-step dewetting further increases the nanoisland diameters and the packing density up to 40%. These morphological modifications induce a redshift of plasmon resonance (Fig. S3) and enhance E-field intensity for highly-sensitive SERS detection. The fluorocarbon thickness is determined by signal-to-noise ratio (SNR) in SERS measurements. Figure 2 b shows the SERS intensity of 10 µM Rhodamine 6G (R6G) at 1365 cm⁻¹ and SNR depending on the fluorocarbon thickness. A thick fluorocarbon layer enhances the SERS intensity (Fig. S4a) but introduces SERS background noise (Fig. S4b). Therefore, the fluorocarbon thickness was optimized to 2 nm to maximize the SNR, defined as the ratio of the SERS peak intensity at 1365 cm⁻¹ to the standard deviation of the fluorocarbon SERS noise signals. Next, the Ag film thickness is optimized for the maximum SERS intensity. Figure 2 c shows SERS intensity for 1 µM R6G at 1365 cm⁻¹ depending on the Ag film thickness and dewetting repetition. A thick Ag film facilitates the formation of large, densely packed nanoislands during the dewetting process (Fig. S5), leading to enhanced SERS intensity (blue bar). Repeated dewetting further enhances the SERS intensity (red bar) by providing abundant plasmonic hotspots. Note that the coalescence of nanoislands at a total thickness of 24 nm leads to a drastic decline in the SERS signal (Fig. S6). As a result, the CEP-SERS patch is fabricated by a 10 nm repeated dewetting process, which provides the maximum SERS intensity (Fig. S7). The SERS intensity of R6G exhibits strong linearity with concentration (Fig. 2 d and Fig. S8). The SERS substrate provides an average SERS enhancement factor of 1.8 × 10 7 with a uniformity of 11.8% (Fig. S9). The CEP-SERS patches are vacuum-sealed for further on-body analysis (Fig. S10) and over 85% of the performance is preserved for up to 25 days (Fig. 2 e). The mechanical stability of the CEP-SERS patch is confirmed by measuring the SERS intensity of benzenethiol at 1070 cm⁻¹. The SERS performance remains stable with no notable degradation after 200 cycles of twisting (red) and bending (blue) (Fig. 2 f). The CEP-SERS patch remains stable and effective for wearable sweat profiling. Microfluidic Sequential Sampler The sequential sweat sampler allows chronological sample collection through capillary bursting valves (CBVs) in a microfluidic network. Figure 3 a illustrates the sampling sequence (left) and SEM images of the microfabricated CBVs (right). The bursting pressure (BP) gradients guide a sample fluid into different chambers for sequential sampling. The BPs are controlled by the width and diverging angle of the CBVs. In particular, the narrow width and the high diverging angle increase the BP in the hydrophobic channel (Fig. S11). The sample fluid enters the first chamber through CBV 1 and fills the chamber. Once the chamber is filled, the fluid then flows into the subsequent chamber through CBV 2. CBV 3 has the highest BP for air ventilation during the sampling. Note that air ventilation allows stable sweat collection without the formation of bubbles. The microfluidic parameters are determined by two-phase fluid dynamic analysis (COMSOL Multiphysics 6.1, Fig. S12). The sequential loading of distinct colored solutions visualizes the stepwise procedure of chrono-sample collection (Fig. 3 b). In addition, the sampling interval is determined by the chamber volume and the flow rate (Fig. 3 c). Each chamber sequentially stores a collected sample in a separate space, and the small chamber volume allows dense sampling intervals over time (Fig. S13 and Supplementary Video 1). In this experiment, the sweat patch with a microchamber volume of 0.5 µL fulfills each chamber within ~ 2 minutes at a flow rate of 0.25 µL/min. The collected samples are preserved without mixing through on-chip isolation and evaporation suppression. The isolation prevents diffusion between fluids, allowing for long-term storage without mixing (Fig. 3 d). An air pocket barrier creates trapped microbubbles that separate each sample. The effectiveness of isolation is demonstrated by measuring SERS intensity from the deionized water (DI water) in the chamber after sequentially injecting DI water and R6G solution. The isolated chambers maintain initial concentration for 30 hours (red), while the absence of the air pocket barrier leads to mixing (blue). Low-permeable polyester film, attached to the medical adhesive at the sequential sampler, prevents evaporation. This method retains about 80% of samples after 30 hours. Note that evaporation-induced loss is quantified by measuring fluid and air bubble fractions (Fig. S14). This microfluidic sampler facilitates stable sweat collection on the skin and long-term preservation. Machine-learned label-free quantification of metabolites The CEP-SERS patch allows label-free quantification of assorted metabolites. Figure 4 a shows the SERS peak intensities of uric acid (red, at 635 cm -1 ), lactate (blue, at 859 cm -1 ), and tyrosine (green, at 1353 cm -1 ) depending on the different concentrations in the range of physiological level (SERS spectra are presented in Fig. S15). A strong linear relationship between SERS peak intensity and concentration facilitates label-free metabolite quantification. Note that intrinsic Raman background noises from PDMS were initially removed for accurate signal interpretation of metabolites. The metabolic profiling over time is further shown by tracking the changes in the SERS peak intensity of the target molecules within artificial sweat samples (Fig. S16a). The uric acid concentration is sequentially adjusted to 80 µM, 10 µM, and 20 µM over the injection time by using a syringe pump (Fig. S16b). Figure 4 b illustrates the variation in SERS peak intensity of uric acid in chrono-sampled artificial sweat over the collection period (left), utilizing the CEP-SERS patch with chamber volumes of 1.5 µL (blue) and 0.5 µL (red) (SERS spectra are presented in Fig. S16c and d). The CEP-SERS patch with a small chamber volume effectively improves the precise quantification of rapid variations in target molecular concentration by providing dense sampling intervals (right). Machine-learned label-free quantification of metabolites is further conducted by using an autoencoder with a logistic regression model. The autoencoder-based prediction model is trained to minimize the total loss function, including the reconstruction and prediction losses (Fig. S17) [68]. Target metabolites including uric acid, lactate, and tyrosine are mixed with 41 different combinations of concentration, and a total of 1,476 spectra are provided for robust machine-learned quantification, which considers various background states of sweat (Fig. S18 and S19). The SERS spectra mapped onto the two-dimensional (2D) latent space via autoencoder show that the individual classes are diagonally aligned depending on the concentration of each metabolite (Fig. S20). Figure 4 c-e shows the predicted concentrations and the corresponding true values for uric acid (R 2 : 0.71–0.80), lactate (R 2 : 0.65–0.83), and tyrosine (R 2 : 0.82–0.92), which are evaluated with 10 times of repeated random sampling cross-validation (Fig. S21). The concentration of each metabolite is well predicted to the true value and the quantification process is explained through the feature extraction calculated by the Shapley additive explanation (SHAP) value (Fig. S22) [69,70]. The SHAP feature importance (solid line) is presented with corresponding SERS spectra at 10 mM concentration (dotted line) for uric acid (Fig. 4 f), lactate (Fig. 4 g), and tyrosine (Fig. 4 h). The extracted features contain the matched characteristics to the SERS spectra including SERS peaks (colored bar) and notches (gray bar). This validation demonstrates the machine-learned quantifications are explainable and established based on the molecule-specific SERS signals. Label-free human sweat profiling of assorted metabolites The CEP-SERS patch collects human sweat on the skin and label-freely profiles the transient alteration of metabolites. The on-body evaluation includes a treadmill warm-up followed by a climb mill exercise, performed on separate days under fasting conditions and after a purine-rich diet intake (Fig. 5 a, Fig. S23). The CEP-SERS patches are attached to multiple sites on the forehead and shoulder of participants' skin to collect exercise-induced sweat (Fig. 5 b). The sweat samples are collected during the evaluation through the CEP-SERS patch (Fig. 5 c). The inset images show the sweat collection using the patch without plasmonic structures to clearly present the sequential sampling. All flexibility of the patch ensures conformal dermal contact, resulting in stable sequential sweat collection (top). In addition, the CEP-SERS patch securely preserves sweat samples under various physical stresses that may occur in a wearable environment such as compressing, twisting, or detaching (bottom). Note that the biophysical signals such as heart rate, respiratory exchange ratio (RER), oxygen, and carbon dioxide intake are simultaneously measured during the evaluation to monitor the exercise intensity. The CEP-SERS patch performs label-free profiling of chrono-sampled human sweat using machine-learned quantification models for individual metabolites. Sweat samples from four healthy participants are analyzed to assess metabolic differences during exercise with a purine-rich meal versus fasting. The average SERS signals measured from human sweat capture key features of various metabolites in sweat (Fig. S24). SERS signals from chrono-sampled sweat are obtained seven times from each chamber to reduce measurement errors (Fig. S25 - S28). The sweat flow rate for each participant is determined by dividing the total volume of the collected samples by the sampling time (Fig. S29). Label-free uric acid, lactate, and tyrosine quantifications in human sweat are validated using commercial fluorometric assay (FMA) or colorimetric assay (CMA) kits. The sweat is further collected by swiping the microtube to the forehead for FMA and CMA. Machine-learned predictions for uric acid and lactate demonstrate strong agreement with FMA measurements, achieving R² values of 0.96 and 0.86, respectively (Fig. 5 d and e). In addition, tyrosine prediction from participant 2–4 shows an error margin of ~ 5 µM to the CMA kit (Fig. 5 f). For Participant 1, contaminants including skin lipids during sample collection significantly affected the absorbance properties, resulting in a substantial difference of ~ 25 µM compared to the CMA measurement. The metabolites in the collected sweat samples (Fig. 5 g-i) are chronologically profiled with biophysical signals (Fig. 5 j) through the machine-learned SERS quantification models (Chronological profiling results of metabolites and biophysical signals for four participants under fasting conditions and purine-rich diet intake are presented in Fig. S30 - S33). The CEP-SERS patch captures the metabolic alterations associated with the metabolic pathway of each metabolite and dilution. The metabolites in sweat are often diluted over time during perspiration [24,31,34,71]. In contrast, lactate temporarily increases, which reflects the eccrine sweat gland metabolism [24,26,72] and anaerobic metabolism [73,74] during exercise. The dietary intake increases the overall concentrations of uric acid [31] and tyrosine (amino acid) [30,71,75] in the on-body evaluation (Fig. 5 k and 5 l). The metabolic alterations under fasting (blue) and purine-rich diet intake conditions (red) show physiological states for four participants. The averaged alteration (bold red line) reveals that lactate levels remain stable due to minimal dietary influence. In contrast, uric acid and tyrosine increase post-ingestion, reflecting physiological changes in purine metabolism and protein digestion, respectively (Fig. 5 m). The CEP-SERS patch combined with machine-learned metabolic quantification has successfully captured transient physiological changes by label-free and multiplexed detection of exercise- and intake-induced metabolic alterations. Conclusions Metabolic profiling offers significant insights into individual physiological states, enhancing the scope of digital phenotyping and advancing personalized healthcare. Wearable sweat sensors non-invasively capture the transient biochemical alterations during assorted activities and allow on-body metabolic profiling. Furthermore, SERS holds significant potential to unveil comprehensive physiological information by facilitating label-free universal molecular recognition in sweat. Unlike others (Table S1 ), our CEP-SERS patch features the integration of nanoplasmonics and functional microfluidics for precise chronological profiling of metabolites. The plasmonic nanostructures are integrated into the microfluidic sampler via large-area fabrication of Ag nanoislands on fluorocarbon-coated PDMS, facilitating both SERS analysis and stable sequential sweat collection. In addition, SERS spectra from chrono-sampled sweat samples are profiled through a robust machine-learned quantification model that accounts for various concentration combinations of the background in sweat. The CEP-SERS patch has successfully captured the exercise- and intake-induced alterations in multiple metabolites. This CEP-SERS patch combined with machine-learned quantification can provide a new strategy for delineating the physiological phenotype of individuals in personalized healthcare. Methods Numerical analysis E-field distribution. The electric fields of Ag nanoislands with varying geometries were numerically calculated utilizing a three-dimensional finite-difference time-domain method (Lumerical FDTD Solutions). Geometric parameters of the Ag nanoislands were derived from binary segmented SEM images employing Image J software, with Ag nanoislands modeled as a cylindrical shape. Fluid dynamics. Fluid dynamics at the capillary bursting valve were numerically calculated by utilizing finite elements methods (FEM, COMSOL Multiphysics 6.1). The calculations were conducted by integrating the two-phase flow and level-set physics modules. In addition, the two-phase flow, level set, and wetted wall phenomena were coupled to derive comprehensive computational results. A time-dependent solver was utilized to compute the temporal evolution of the sequential sampling. Materials and reagents Rhodamine 6G (R6G, R4127-25G, dye content ~95%), benzenethiol (W361607-SAMPLE-K, ≥98%), sodium L-lactate (L7022-5G, ~98%), urea (U5378-100G), tyrosine (PHR1097-1G, certified reference material), glycine (G7126-100G, ≥99%), L-alanine (A7469-25G, ≥98.5%), L-glutamic acid monosodium salt monohydrate (49621-250G, ≥98.0%), uric acid (U2625-25G, ≥99%), L-ascorbic acid (A5960-25G, ≥99.0%), D-(+)-glucose (G8270-100G, ≥99.5%), creatinine (PHR1462-1G, certified reference material), sodium chloride (S7653-250G, ≥99.5%), and potassium chloride (P5405-250G, ≥ 99.0%) are purchased from Sigma Aldrich. Medical adhesives were purchased from 3M (Tegaderm 1622W, 1522). Fabrication of CEP-SERS patch A Si wafer was immersed in buffered oxide etchant for 30 seconds to remove the native oxide layer. Photoresist (SU-8 2000.5, MicroChem Corp.) was then spin-coated to a thickness of 500 nm to form an adhesion layer. An additional photoresist (SU-8 2100, MicroChem Corp.) was applied to fabricate a microfluidic channel mold with a thickness of 200 µm. A 10:1 mixture of PDMS base and curing agent (Sylgard 184, Dow Corning Corp.) was spin-coated onto the microfluidic channel mold at 100 rpm for 30 seconds and then at 160 rpm for 30 seconds. The PDMS was cured on a hotplate at 80°C for 1 hour and 30 minutes. The PDMS layer was peeled off from the mold after curing and washed in isopropyl alcohol for 5 minutes. The fluorocarbon layer was coated on the microfluidic channel via AP-CVD with parameters set to 150 W plasma power, a helium flow rate of 5.0 L/min, and a fluorocarbon flow rate of 2 sccm (IHP-1000, APP Korea). Ag film was thermally deposited at a deposition rate of 1.0 Å/s by using a thermal evaporator (SNTEC Inc., Korea) and dewetted on a hotplate at 160°C for 30 minutes. The Ag deposition and dewetting steps were repeated twice to create strong plasmonic hotspots. The air ventilation outlet and sweat inlet were punched on PCL and medical adhesive film (Tegaderm 1622W, 3M) for DCL, respectively. The two layers were bonded by double-sided medical adhesive (1522, 3M) to complete the CEP-SERS patch fabrication. Characterization and measurement Absorption measurement. Absorption spectra were measured by using a microscopic spectrometer setup comprising an inverted microscope (Axiovert 200M, Carl Zeiss ) integrated with a white light LED lamp (MCWHL5-C4, Thorlabs Inc.) and a spectrometer (MicroSpec 2300i) featuring a charge-coupled device (CCD) camera (Model PIXIS: 400BR, Princeton Instruments). Absorption spectra of Ag nanoislands on fluorocarbon-coated PDMS were acquired utilizing a 20× objective lens (NA=0.5). SERS measurement. A helium-neon laser (HRP050, Thorlabs Inc.) operating at a wavelength of 633 nm was utilized in conjunction with a spectrometer featuring a CCD camera, both integrated with an inverted microscope. Light excitation and collection were performed via a 20× objective lens (NA = 0.5). The excitation laser was operated at a power of 5 mW, while data acquisition times were set to 1 second for characterization and extended to 10 seconds for metabolites and human sweat sample measurements. Note that a single SERS spectrum from human sweat was acquired by averaging 6 times of the measurement to minimize signal variances. All the SERS spectra are measured as solution state, and an equivalent volume of DI water was applied to the sample before detection for dry samples. Machine-learned metabolic quantification SERS data preparation. First, 1,476 spectra with 41 different concentration combinations including target molecules are provided for machine-learned quantification. The measured SERS spectra have a vector length of 1321, which contains the signal intensities for the Raman shift range of 457–1674 cm –1 . Each vector was normalized to a range of 0-1. Metabolic quantification. The model consists of a symmetric encoder and decoder with four equally sized layers of 1321 nodes, maintaining a consistent number of neurons across all layers, with a latent layer of length 2 between the encoder and decoder. The model was trained with the Adam optimizer, and the hyperparameters were heuristically tuned to optimize performance. Initially, the learning rate, weight decay, and batch size were set to 1e -4 , 1e -5 , and 32 respectively. The maximum number of training epochs was set to 150, with early stopping employed. After the 50 th epoch, a scheduled weight decay adjustment was applied to promote further model generalization, reducing the learning rate and weight decay to 1e -5 and 1e -6 , respectively. The model predicts the concentration of each metabolite based on the values along the concentration axis in the latent space. Note that three-quarters of the entire SERS spectra data were used for training, with the remainder reserved for validation. In addition, 10 times of repeated random sampling cross-validation ensures robust model evaluation. SERS feature importance calculation. SHAP (SHapley Additive exPlanations) values were utilized to identify the significant features for concentration prediction. The SHAP calculations were performed using only the encoder part of the model, focusing on the feature importance independent of the increase or decrease in concentration. For each sample, the SHAP values were squared and then averaged across all samples to obtain the overall importance of each feature, thereby identifying key spectral features contributing to the model's predictions. On-body evaluation Protocols. The participant performed the following exercise protocol on separate days under fasting conditions and after the purine-rich diet intake, which included 125 g to 250 g of sardines. On the days when the purine-rich diet was consumed, the exercise was initiated after a 1.5-hour rest period. The participants wear the CEP-SERS patches attached to the forehead and shoulder, a wearable metabolic system (Cosmed K5, Cosmed) for respiratory gas analysis, and a heart rate sensor (Polar OH1, Polar Electro Oy) for heart rate monitoring. The participant performed a warm-up consisting of two cycles of running on a treadmill at a 7.2 km/hr speed for 4 minutes, followed by a 3-minute rest period. Subsequently, the participant engaged in exercise on a climbmill with progressively increasing intensity until the heart rate reached 80% of the maximum heart rate. The participant completed an additional 5-minute rest period after exercise, concluding the protocol. Note that sweat samples were additionally collected using a microtube swapped on the forehead to measure the actual concentration of sweat using the fluorometric assay or colorimetric assay kit. Sweat was collected once after the treadmill exercise, at 3-5 minute intervals during the climb mill exercise, and once after a 5-minute rest period following the completion of all exercise tasks. All human trials were approved by the Institutional Review Board of Korea Advanced Science and Technology (protocol number: KH2024-085). In addition, we obtained informed consent from all participants before the experiment. Fluorometric assays for validation of uric acid and lactate quantification. The concentrations of uric acid and lactate in collected sweat were validated by using a commercial uric acid assay kit (MAK077-1KT, Sigma Aldrich) and a lactate assay kit (BM-LAC-100, PicoSens TM ). For uric acid measurement, 5 µL of sweat sample was mixed with 45 µL of uric acid assay buffer, followed by the addition of 2.17 µL of probe and 2.17 µL of enzyme mix. The mixture was thoroughly mixed by pipetting and incubated for 30 minutes at 37°C. For lactate measurement, sweat samples were diluted 200-fold with deionized water. A 10 µL aliquot of the diluted sweat sample was mixed with 40 µL of lactate assay buffer, followed by the addition of 0.43 µL of probe and 2.17 µL of enzyme mix. The mixture was thoroughly mixed by pipetting and incubated for 30 minutes at room temperature. Fluorescence intensity was measured at excitation and emission wavelengths of 535 nm and 590 nm, respectively. Uric acid and lactate concentrations were calculated using a standard curve of each molecule. Colorimetric assays for validation of tyrosine quantification. The concentrations of tyrosine in collected sweat were validated by using a commercial tyrosine assay kit (MET-5073, Cell Biolabs Inc.). Sweat samples were prepared by adjusting their volume to 50 µL with standard tyrosine solutions. The mixed samples were chilled on ice for 30 minutes and then centrifuged at 10,000 g for 15 minutes at 4°C. The supernatant was carefully transferred to prevent the pellet from dissolving and subsequently passed through a QIAquick Spin Column (28115) by centrifugation (MICRO 17 TR, Hanil Science Co.) at 5,000 g for 10 minutes at 4 °C. The filtered samples were then used for tyrosine quantification. The 10x enzyme from the kit was diluted to the assay solution, and 15 µL of each prepared sample and standard tyrosine solution was added to individual wells of a transparent 384-well plate, followed by 15 µL of the diluted enzyme solution to make a total volume of 30 µL per well. The plate was incubated at room temperature for 10 minutes on a horizontal shaker, and absorbance was measured at 490 nm using a plate reader (CYTATION 5 imaging reader, Agilent BioTek). Tyrosine concentrations were then calculated using a standard curve. Declarations Disclosures The authors declare no conflicts of interest. Data availability The authors declare that the main data supporting the findings of this study are available within the article and its Supplementary Information file. The accession codes for raw SERS spectra will be available before publication. Code availability Code underlying the results presented in this paper is not publicly available at this time but may be obtained from the authors upon request. Acknowledgments. This work was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science ICT & Future Planning (2022M3H4A4085645, RS-2024-00438316), the Technology Innovation Program (No. RS-2024-00432381) funded by the Ministry of Trade Industry & Energy (MOTIE, Korea). Authors information Authors and Affiliations Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea Jaehun Jeon, Sangyeon Lee, Hanjin Kim, Eun-Sil Yu, Hamin Na, Doheon Lee & Ki-Hun Jeong KAIST Institute for Health Science and Technology (KIHST), KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon, Republic of Korea, 305-701 Jaehun Jeon, Eun-Sil Yu, Hamin Na & Ki-Hun Jeong Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon, 34141, Korea Seongok Chae & Hyung-Soon Park Bionanotechnology Research Center, Korea Research Institute of Bioscience and Biotechnology (KRIBB), 125 Gwahak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea Joo Hoon Lee & Taejoon Kang Contributions J.J. and K.-H.J. conceived the idea and designed the projects. K.-H.J. supervised the overall research. J.J. designed, fabricated, characterized, and packaged the all-flexible chronoepifluidic SERS patch and conducted overall experiments. S.L., H.K., and D.L. contributed to the machine-learned quantification of metabolites. J.J., S.C., and H.-S.P. conducted the on-body evaluation of the SERS patch. J.J., J.H.L., and T.K. validated the SERS patch using the commercial fluorometric and colorimetric kit. J.J., K.-H.J., S.L., S.C., J.H.L., H.K., E.-S.Y., H.N., T.K., H.-S.P., and D.L. evaluated the experiments and contributed valuable ideas. J.J. and K.-H.J. wrote the manuscript. All authors discussed the results and commented on the manuscript. Corresponding author Correspondence to Ki-Hun Jeong. References Holmes, E., Wilson, I. D., and Nicholson, J. K. Metabolic phenotyping in health and disease. Cell, 134(5), 714–717 (2008). Holmes, E., Loo, R. L., Stamler, J., Bictash, M., Yap, I. K., Chan, Q., … and Elliott, P. Human metabolic phenotype diversity and its association with diet and blood pressure. Nature, 453(7193), 396–400 (2008). Nicholson, J. K., Holmes, E., Kinross, J. M., Darzi, A. W., Takats, Z., and Lindon, J. C. Metabolic phenotyping in clinical and surgical environments. Nature, 491(7424), 384–392 (2012). Dumas, M. E., Kinross, J., and Nicholson, J. K. Metabolic phenotyping and systems biology approaches to understanding metabolic syndrome and fatty liver disease. Gastroenterology, 146(1), 46–62 (2014). Holmes, E., Wijeyesekera, A., Taylor-Robinson, S. D., and Nicholson, J. K. The promise of metabolic phenotyping in gastroenterology and hepatology. Nature reviews Gastroenterology & hepatology, 12(8), 458–471 (2015). Everett, J. R., Holmes, E., Veselkov, K. A., Lindon, J. C., and Nicholson, J. K. A unified conceptual framework for metabolic phenotyping in diagnosis and prognosis. Trends in Pharmacological Sciences, 40(10), 763–773 (2019). Suhre, K. Metabolic profiling in diabetes. J Endocrinol, 221(3), R75-R85 (2014). Inoue, K., Tsutsui, H., Akatsu, H., Hashizume, Y., Matsukawa, N., Yamamoto, T., and Toyo'Oka, T. Metabolic profiling of Alzheimer's disease brains. Scientific reports, 3(1), 2364 (2013). Secor, S. M. Specific dynamic action: a review of the postprandial metabolic response. Journal of Comparative Physiology B, 179, 1–56 (2009). Emerson, S. R., Kurti, S. P., Emerson, E. M., Cull, B. J., Casey, K., Haub, M. D., and Rosenkranz, S. K. Postprandial metabolic responses differ by age group and physical activity level. The Journal of nutrition, health and aging, 22(1), 145–153 (2018). Fiehn, O. and Kind, T. Metabolite profiling in blood plasma. Metabolomics: Methods and protocols, 3–17 (2007). Grauslys, A., Phelan, M. M., Broughton, C., Baines, P. B., Jennings, R., Siner, S., … and Carrol, E. D. NMR-based metabolic profiling provides diagnostic and prognostic information in critically ill children with suspected infection. Scientific Reports, 10(1), 20198 (2020). Want, E., Wilson, I., Gika, H. et al. Global metabolic profiling procedures for urine using UPLC–MS. Nature Protocols 5, 1005–1018 (2010). Mathé, E. A., Patterson, A. D., Haznadar, M., Manna, S. K., Krausz, K. W., Bowman, E. D., … and Harris, C. C. Noninvasive urinary metabolomic profiling identifies diagnostic and prognostic markers in lung cancer. Cancer research, 74(12), 3259–3270 (2014). Luan, H., Liu, L. F., Tang, Z., Zhang, M., Chua, K. K., Song, J. X., … and Cai, Z. Comprehensive urinary metabolomic profiling and identification of potential noninvasive marker for idiopathic Parkinson’s disease. Scientific reports, 5(1), 13888 (2015). Park, W., Seo, H., Kim, J. et al. In-depth correlation analysis between tear glucose and blood glucose using a wireless smart contact lens. Nature Communications 15, 2828 (2024). Keum, D. H., Kim, S. K., Koo, J., Lee, G. H., Jeon, C., Mok, J. W., … and Hahn, S. K. Wireless smart contact lens for diabetic diagnosis and therapy. Science Advances, 6(17), eaba3252 (2020). Kim, J., Imani, S., de Araujo, W. R., Warchall, J., Valdés-Ramírez, G., Paixão, T. R., … and Wang, J. Wearable salivary uric acid mouthguard biosensor with integrated wireless electronics. Biosensors and Bioelectronics, 74, 1061–1068 (2015). Brunmair, J., Gotsmy, M., Niederstaetter, L., Neuditschko, B., Bileck, A., Slany, A., … and Gerner, C. Finger sweat analysis enables short interval metabolic biomonitoring in humans. Nature Communications, 12(1), 5993 (2021). Bariya, M., Li, L., Ghattamaneni, R., Ahn, C. H., Nyein, H. Y. Y., Tai, L. C., and Javey, A. Glove-based sensors for multimodal monitoring of natural sweat. Science advances, 6(35), eabb8308 (2020). Tu, J., Torrente-Rodríguez, R. M., Wang, M. and Gao, W. The era of digital health: a review of portable and wearable affinity biosensors. Advanced Functional Materials, 30, 1906713 (2020). Heikenfeld, J. et al. Accessing analytes in biofluids for peripheral biochemical monitoring. Nature Biotechnology, 37, 407–419 (2019). D. S. Yang, R. Ghaffari, and J. A. Rogers Sweat as a diagnostic biofluid. Science, 379(6634), 760–761 (2023). Min, J., Tu, J., Xu, C., Lukas, H., Shin, S., Yang, Y., … and Gao, W. Skin-interfaced wearable sweat sensors for precision medicine. Chemical reviews, 123(8), 5049–5138 (2023). Yang, G., Hong, J., and Park, S. B. Wearable device for continuous sweat lactate monitoring in sports: a narrative review. Frontiers in Physiology, 15, 1376801 (2024). Komkova, M. A., Eliseev, A. A., Poyarkov, A. A., Daboss, E. V., Evdokimov, P. V., Eliseev, A. A., & Karyakin, A. A. Simultaneous monitoring of sweat lactate content and sweat secretion rate by wearable remote biosensors. Biosensors and Bioelectronics, 202, 113970 (2022). Xuan, X., Chen, C., Molinero-Fernandez, A., Ekelund, E., Cardinale, D., Swarén, M., … and Crespo, G. A. Fully integrated wearable device for continuous sweat lactate monitoring in sports. ACS sensors, 8(6), 2401–2409 (2023). Lynch, C., Adams, S. Branched-chain amino acids in metabolic signaling and insulin resistance. Nature Reviews Endocrinology 10, 723–736 (2014). Mayers, J., Wu, C., Clish, C. et al. Elevation of circulating branched-chain amino acids is an early event in human pancreatic adenocarcinoma development. Nature Medicine 20, 1193–1198 (2014). M. Wang, Y. Yang, J. Min, Y. Song, J. Tu, D. Mukasa, … and W. Gao A wearable electrochemical biosensor for the monitoring of metabolites and nutrients. Nature Biomedical Engineering, 6(11), 1225–1235 (2022). Yang, Y., Song, Y., Bo, X. et al. A laser-engraved wearable sensor for sensitive detection of uric acid and tyrosine in sweat. Nature Biotechnology 38, 217–224 (2020). Russell, E., Koren, G., Rieder, M., and Van Uum, S. H. The detection of cortisol in human sweat: implications for measurement of cortisol in hair. Therapeutic drug monitoring, 36(1), 30–34 (2014). Parlak, O., Keene, S. T., Marais, A., Curto, V. F., and Salleo, A. Molecularly selective nanoporous membrane-based wearable organic electrochemical device for noninvasive cortisol sensing. Science advances, 4(7), eaar2904 (2018). Tang, W., Yin, L., Sempionatto, J. R., Moon, J. M., Teymourian, H., and Wang, J. Touch-based stressless cortisol sensing. Advanced Materials, 33(18), 2008465 (2021). Parrilla, M., Guinovart, T., Ferré, J., Blondeau, P., and Andrade, F. J. A wearable paper-based sweat sensor for human perspiration monitoring. Advanced Healthcare Materials, 8(16), 1900342 (2019). Nagamine, K., Mano, T., Nomura, A., Ichimura, Y., Izawa, R., Furusawa, H., … and Tokito, S. Noninvasive sweat-lactate biosensor emplsoying a hydrogel-based touch pad. Scientific reports, 9(1), 10102 (2019). Zhao, F. J., Bonmarin, M., Chen, Z. C., Larson, M., Fay, D., Runnoe, D., and Heikenfeld, J. Ultra-simple wearable local sweat volume monitoring patch based on swellable hydrogels. Lab on a Chip, 20(1), 168–174 (2020). Nyein, H.Y.Y., Bariya, M., Tran, B. et al. A wearable patch for continuous analysis of thermoregulatory sweat at rest. Nature Communications 12, 1823 (2021). Reeder, J.T., Xue, Y., Franklin, D. et al. Resettable skin interfaced microfluidic sweat collection devices with chemesthetic hydration feedback. Nature Communications, 10, 5513 (2019). Reeder, J. T., Choi, J., Xue, Y., Gutruf, P., Hanson, J., Liu, M., … and Rogers, J. A. Waterproof, electronics-enabled, epidermal microfluidic devices for sweat collection, biomarker analysis, and thermography in aquatic settings. Science Advances, 5(1), eaau6356 (2019). Gao, W., Emaminejad, S., Nyein, H. et al. Fully integrated wearable sensor arrays for multiplexed in situ perspiration analysis. Nature 529, 509–514 (2016). S. Emaminejad, W. Gao, W. Wu, Z. A. Davies, H. Yin Yin Nyein, S. Challa, … and R. W. Davis Autonomous sweat extraction and analysis applied to cystic fibrosis and glucose monitoring using a fully integrated wearable platform. Proceedings of the National Academy of sciences , 114(18), 4625–4630 (2017). R. T. Ray, M. Ivanovic, P. M. Curtis, D. Franklin, K. Guventurk, W. J. Jeang, … and J. A. Rogers Soft, skin-interfaced sweat stickers for cystic fibrosis diagnosis and management. Science translational medicine, 13(587), eabd8109 (2021). Koh, A., Kang, D., Xue, Y., Lee, S., Pielak, R. M., Kim, J., … and Rogers, J. A. A soft, wearable microfluidic device for the capture, storage, and colorimetric sensing of sweat. Science translational medicine, 8(366), 366ra165-366ra165 (2016). Kim, J., Wu, Y., Luan, H., Yang, D. S., Cho, D., Kwak, S. S., … and Rogers, J. A. A skin-interfaced, miniaturized microfluidic analysis and delivery system for colorimetric measurements of nutrients in sweat and supply of vitamins through the skin. Advanced Science, 9(2), 2103331 (2022). Wang, Y., Zhao, C., Wang, J., Luo, X., Xie, L., Zhan, S., … and Ying, Y. Wearable plasmonic-metasurface sensor for noninvasive and universal molecular fingerprint detection on biointerfaces. Science Advances, 7(4), eabe4553 (2021). Bi, X., Czajkowsky, D.M., Shao, Z. et al. Digital colloid-enhanced Raman spectroscopy by single-molecule counting. Nature (2024). Yu, W., Li, Q., Ren, J., Feng, K., Gong, J., Li, Z., … and Yang, L. A sensor platform based on SERS detection/janus textile for sweat glucose and lactate analysis toward portable monitoring of wellness status. Biosensors and Bioelectronics, 263, 116612 (2024). Ma, H., Cui, Q., Xu, L., Tian, Y., Jiao, A., Wang, C., … and Chen, M. Silk fibroin fibers decorated with urchin-like Au/Ag nanoalloys: a flexible hygroscopic SERS sensor for monitoring of folic acid in human sweat. Optics Express, 29(19), 30892–30904 (2021). Zhang, X., Wang, X., Ning, M., Wang, P., Wang, W., Zhang, X., … and Li, S. Fast synthesis of Au nanoparticles on metal–phenolic network for sweat SERS analysis. Nanomaterials, 12(17), 2977 (2022). Koh, E. H., Lee, W. C., Choi, Y. J., Moon, J. I., Jang, J., Park, S. G., … and Jung, H. S. A wearable surface-enhanced Raman scattering sensor for label-free molecular detection. ACS Applied Materials and Interfaces, 13(2), 3024–3032 (2021). Chen, Z., Wang, W., Tian, H., Yu, W., Niu, Y., Zheng, X., … and Huang, Y. Wearable intelligent sweat platform for SERS-AI diagnosis of gout. Lab on a Chip, 24(7), 1996–2004 (2024). Ye, H., Chen, X., Huang, X., Li, C., Yin, X., Zhao, W., and Wang, T. Patterned Gold Nanoparticle Superlattice Film for Wearable Sweat Sensors. Nano Letters, 24 (35), 11082–11089 (2024). Xiong, S., Wang, C., Zhu, C., Dong, P., and Wu, X. Dual Detection of Urea and Glucose in Sweat Using a Portable Microfluidic SERS Sensor with Silver Nano-Tripods and 1D-CNN Model Analysis. ACS Applied Materials & Interfaces (2024). Hu, M., Zhu, K., Wei, J., Yang, K., Wu, L., Zong, S., and Wang, Z. Silk fibroin-based wearable SERS biosensor for simultaneous sweat monitoring of creatinine and uric acid. Biosensors and Bioelectronics, 265 , 116662 (2024). Guan, P. C., Qi, Q. J., Wang, Y. Q., Lin, J. S., Zhang, Y. J., and Li, J. F. Development of a 3D Hydrogel SERS Chip for Noninvasive, Real-Time pH and Glucose Monitoring in Sweat. ACS Applied Materials & Interfaces, 16 (36), 48139–48146 (2024). Zhang, H., Zhang, H., Sikdar, D., Liu, X., Yang, Z., Cheng, W., and Chen, Y. Jellyfish-like Gold Nanowires as FlexoSERS Sensors for Sweat Analysis. Nano Letters, 24 (36), 11269–11278 (2024). He, X., Fan, C., Luo, Y., Xu, T., and Zhang, X. Flexible microfluidic nanoplasmonic sensors for refreshable and portable recognition of sweat biochemical fingerprint. npj Flexible Electronics, 6(1), 60 (2022). Mogera, U., Guo, H., Namkoong, M., Rahman, M. S., Nguyen, T., and Tian, L. Wearable plasmonic paper–based microfluidics for continuous sweat analysis. Science advances, 8(12), eabn1736 (2022). Li, G., Zhao, X., Tang, X., Yao, L., Li, W., Wang, J.,and Hao, Q. Wearable Hydrogel SERS Chip Utilizing Plasmonic Trimers for Uric Acid Analysis in Sweat. Nano Letters, 24 (42), 13447–13454 (2024). Xiao, J., Wang, J., Luo, Y., Xu, T., and Zhang, X. Wearable plasmonic sweat biosensor for acetaminophen drug monitoring . ACS sensors, 8(4), 1766–1773 (2023). Jain, V., Ochoa, M., Jiang, H. et al. A mass-customizable dermal patch with discrete colorimetric indicators for personalized sweat rate quantification. Microsystems and Nanoengineering 5, 29 (2019). Cheng, Y., Feng, S., Ning, Q. et al. Dual-signal readout paper-based wearable biosensor with a 3D origami structure for multiplexed analyte detection in sweat. Microsystems and Nanoengineering 9, 36 (2023). Li, Y., Guo, Y., Chen, H., Xiao, X., Long, F., Zhong, H., … and Liu, Z. Flexible Wearable Plasmonic Paper-Based Microfluidics with Expandable Channel and Adjustable Flow Rate for Portable Surface-Enhanced Raman Scattering Sweat Sensing. ACS Photonics (2024). Choi, J., Kang, D., Han, S., Kim, S. B., and Rogers, J. A. Thin, soft, skin-mounted microfluidic networks with capillary burst valves for chrono‐sampling of sweat. Advanced healthcare materials, 6(5), 1601355 (2017). Wu, C. H., Ma, H. J. H., Baessler, P., Balanay, R. K., and Ray, T. R. Skin-interfaced microfluidic systems with spatially engineered 3D fluidics for sweat capture and analysis. Science Advances, 9(18), eadg4272 (2023). Wu, Y., Kanatzidis, E. E., Avila, R., Zhou, M., Bai, Y., Chen, S., … and Rogers, J. A. 3D-printed epidermal sweat microfluidic systems with integrated microcuvettes for precise spectroscopic and fluorometric biochemical assays. Materials Horizons, 10(11), 4992–5003 (2023). Lee, S., Kim, H., and Lee, D. Linearization autoencoder: an autoencoder-based regression model with latent space linearization. bioRxiv, 2022–06 (2022). Lundberg, S. M., Erion, G., Chen, H., DeGrave, A., Prutkin, J. M., Nair, B., … and Lee, S. I. From local explanations to global understanding with explainable AI for trees. Nature machine intelligence, 2(1), 56–67 (2020). Chen, H., Covert, I. C., Lundberg, S. M., & Lee, S. I. Algorithms to estimate Shapley value feature attributions. Nature Machine Intelligence, 5(6), 590–601 (2023). Zhong, B., Qin, X., Xu, H., Liu, L., Li, L., Li, Z., … and Wang, L. Interindividual-and blood-correlated sweat phenylalanine multimodal analytical biochips for tracking exercise metabolism. Nature Communications, 15(1), 624 (2024). Derbyshire, P. J., Barr, H., Davis, F., and Higson, S. P. Lactate in human sweat: a critical review of research to the present day. The journal of physiological sciences, 62, 429–440 (2012). Xuan, X., Perez-Rafols, C., Chen, C., Cuartero, M., and Crespo, G. A. Lactate biosensing for reliable on-body sweat analysis. ACS sensors, 6(7), 2763–2771 (2021). Karpova, E. V., Laptev, A. I., Andreev, E. A., Karyakina, E. E., and Karyakin, A. A. Relationship between sweat and blood lactate levels during exhaustive physical exercise. ChemElectroChem, 7(1), 191–194 (2020). Fernstrom, M. H. and Fernstrom, J. D. Protein consumption increases tyrosine concentration and in vivo tyrosine hydroxylation rate in the light-adapted rat retina. Brain research, 401(2), 392–396 (1987). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryVideo1.mp4 Supplementary video 1 SupplementaryInformationKHJeongXKAIST.docx Supplementary information updated Cite Share Download PDF Status: Published Journal Publication published 27 Aug, 2025 Read the published version in Nature Communications → 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-5624954","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":414084691,"identity":"1df83c8c-16a1-4b9e-8319-f1228b3050ac","order_by":0,"name":"Ki-Hun Jeong","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4UlEQVRIie2RsQrCMBCG/yLoEnVNCdRXaAl0Kn0WpeDsAzjERRdxVgSfJRLQpdhVcLE4uDoVkSJGEUfTUTDfcOGOfPAfB1gsPwiFfD6Rj9pnUk3pPxVHVFReKF+Xioo7UuvT4Jbx9qRxPmIYw13K7wrDJuGL2SGkigRCN2Ct7nfFQxqy5vQQQREdrC7hEUMwD1mhlV3UUY1c4F5BYUjrjFxl6CsEwhlLMJPiig3nTZHwQO8y780S4k4NCt2r/ETKOFhl2+PlUsQeTQ0KqNQHGb+bLmCKpWkLXUrzP4vFYvljHpgfPzqOYu/JAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-4799-7816","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Ki-Hun","middleName":"","lastName":"Jeong","suffix":""},{"id":414084692,"identity":"c5d4249d-f676-4345-95e4-7489350ee75f","order_by":1,"name":"Jaehun Jeon","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Jaehun","middleName":"","lastName":"Jeon","suffix":""},{"id":414084693,"identity":"334720a7-b2d9-4aa1-ad7d-435d01b8c249","order_by":2,"name":"Sangyeon Lee","email":"","orcid":"https://orcid.org/0000-0002-3260-4285","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Sangyeon","middleName":"","lastName":"Lee","suffix":""},{"id":414084694,"identity":"6d9e8a57-f0f3-4aa1-9c56-b38078a6575b","order_by":3,"name":"Seongok Chae","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Seongok","middleName":"","lastName":"Chae","suffix":""},{"id":414084695,"identity":"a0a97a7a-31d0-4c05-ba6e-2034bbd8045d","order_by":4,"name":"Joo Hoon Lee","email":"","orcid":"","institution":"Korea Research Institute of Bioscience and Biotechnology","correspondingAuthor":false,"prefix":"","firstName":"Joo","middleName":"Hoon","lastName":"Lee","suffix":""},{"id":414084696,"identity":"62579114-dd94-4e4e-ae81-d3d95ca905dd","order_by":5,"name":"Hanjin Kim","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Hanjin","middleName":"","lastName":"Kim","suffix":""},{"id":414084697,"identity":"efbb970f-0b82-4691-bac9-b34829cf106d","order_by":6,"name":"Eun-Sil Yu","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Eun-Sil","middleName":"","lastName":"Yu","suffix":""},{"id":414084698,"identity":"b4b4ef91-305a-4374-ba51-ba75c6d6f6f4","order_by":7,"name":"Hamin Na","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Hamin","middleName":"","lastName":"Na","suffix":""},{"id":414084699,"identity":"25b0a53f-1ffd-4b0f-aec4-4a2e6c742d76","order_by":8,"name":"Taejoon Kang","email":"","orcid":"https://orcid.org/0000-0002-5387-6458","institution":"KRIBB","correspondingAuthor":false,"prefix":"","firstName":"Taejoon","middleName":"","lastName":"Kang","suffix":""},{"id":414084700,"identity":"f7f2101c-0eee-4d86-b413-4bdb37005d27","order_by":9,"name":"Hyung-Soon Park","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Hyung-Soon","middleName":"","lastName":"Park","suffix":""},{"id":414084701,"identity":"922f7b06-614f-4467-a37b-818dc694adad","order_by":10,"name":"Doheon Lee","email":"","orcid":"","institution":"Korea Advanced Institute of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Doheon","middleName":"","lastName":"Lee","suffix":""}],"badges":[],"createdAt":"2024-12-11 13:56:54","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5624954/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5624954/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-025-63510-2","type":"published","date":"2025-08-27T04:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":76088290,"identity":"8ffe999c-e4ce-4029-86ba-c45ed1b45c25","added_by":"auto","created_at":"2025-02-12 08:04:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2976975,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAll-flexible chronoepifluidic SERS patch.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Schematic illustration of an all-flexible chronoepifluidic SERS patch (CEP-SERS patch) comprising plasmofluidic channel layer (PCL) and dermal-contact layer (DCL). PCL includes an all-flexible plasmonic SERS substrate and a chronoepifluidic sweat sampler. The flexible SERS substrate features plasmonic nanoislands on ultrathin fluorocarbon-coated PDMS membrane using low-temperature solid-state dewetting of thin silver film. The sweat sampler sequentially collects the sweat via the microfluidic channel with capillary bursting valves, which spatially separates the sweat over time. DCL interconnects the PCL on the skin for stable sweat collection, which contains medical adhesive with a sweat collection port. The CEP-SERS patch provides conformal contact on human skin and label-free sweat profiling of diverse metabolites from sequentially sampled sweat. (\u003cstrong\u003eb\u003c/strong\u003e) Schematic illustration of micro and nanofabrication for the CEP-SERS patch. The device fabrication includes fluorocarbon coating (S1), thermal evaporation of Ag (S2), low-temperature solid-state dewetting (S3), and microfluidic encapsulation (S4). The ultrathin fluorocarbon film effectively dewets the thin metal film on the sweat sampler, resulting in plasmonic structures with strong electromagnetic hotspots, which facilitate highly sensitive SERS analysis. Optical images of (\u003cstrong\u003ec\u003c/strong\u003e) the PCL (Scale bar: 1 mm) with an inset SEM image of Ag nanoislands (Scale bar: 100 nm) and (\u003cstrong\u003ed\u003c/strong\u003e) the CEP-SERS patch on the skin (Scale bar: 10 mm).\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/79d2bc94c9eabd7f072ed0c2.png"},{"id":76088292,"identity":"41750617-73b2-4e40-b650-e6e03cf0d1c1","added_by":"auto","created_at":"2025-02-12 08:04:03","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":2648492,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAll-flexible SERS Substrate: Ag nanoislands on ultrathin fluorocarbon.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) SEM images (left) and E-field distribution (middle) of Ag nanoislands (Scale bar:\u0026nbsp; 100 nm), structural features including diameter and surface coverage (right top), and absorption spectra (right bottom) for 10 nm Ag film on PDMS without fluorocarbon coating (green), single-dewetted 10 nm Ag film (red), and double-dewetted 10 nm Ag film (blue) on fluorocarbon coated PDMS. (\u003cstrong\u003eb\u003c/strong\u003e) SERS intensity of 10 µM R6G at 1365 cm⁻¹ and SNR depending on the fluorocarbon thickness (t\u003csub\u003eFC\u003c/sub\u003e). (\u003cstrong\u003ec\u003c/strong\u003e) SERS intensity for 1 µM R6G depending on the Ag film thickness and dewetting repetition (blue: single dewetting, red: repeated dewetting). (\u003cstrong\u003ed\u003c/strong\u003e) SERS intensity depends on different concentrations of R6G. (\u003cstrong\u003ee\u003c/strong\u003e) Long-term stability of the CEP-SERS patch depending on storage duration measured by SERS peak intensity of 10 μM R6G solution. (\u003cstrong\u003ef\u003c/strong\u003e) Mechanical stability measured by SERS peak intensity variation of benzenethiol at 1070 cm\u003csup\u003e−1\u003c/sup\u003e after 200 cycles of twisting (red) and bending (blue) (Scale bar: 5 mm). The error bars represent one standard deviation from the mean.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/bd4160c6bde95127b7e7e4d1.png"},{"id":76088293,"identity":"e58d701f-9272-44ee-9a37-67b540a84871","added_by":"auto","created_at":"2025-02-12 08:04:03","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":6249670,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eChrono sample collection and isolation through microfluidic sequential sampler.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Schematic illustration of sequential sampling (left) and SEM images of capillary bursting valves on the microfluidic sequential sampler (right) (Scale bar: 200 μm). (\u003cstrong\u003eb\u003c/strong\u003e) Optical images of sequential sampling of colored dye (Scale bar: 5 mm). (\u003cstrong\u003ec\u003c/strong\u003e) Measured sampling interval depending on flow rate and chamber volume. Each color represents varying chamber volume (blue: 1.5 μL, green: 0.9 μL, red: 0.5 μL). (\u003cstrong\u003ed\u003c/strong\u003e) Schematic illustration of on-chip sample isolation utilizing air pocket barrier. (\u003cstrong\u003ee\u003c/strong\u003e) The effectiveness of sample isolation is demonstrated by measuring the SERS intensity of diffused R6G at the chamber filled with DI water depending on storage duration with (red) and without (blue) sample isolation. The error bars represent one standard deviation from the mean.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/665c4091a5059075a4aab9ac.png"},{"id":76088297,"identity":"9d44672e-5b85-4adf-8e6e-37eade845ac0","added_by":"auto","created_at":"2025-02-12 08:04:03","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":4160265,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMachine-learned label-free quantification of metabolites.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) SERS peak intensity of uric acid (red, at 635 cm\u003csup\u003e-1\u003c/sup\u003e), lactate (blue, at 859 cm\u003csup\u003e-1\u003c/sup\u003e), and tyrosine (green, at 1353 cm\u003csup\u003e-1\u003c/sup\u003e) depending on the concentration. The error bars represent one standard deviation from the mean. (\u003cstrong\u003eb\u003c/strong\u003e) Uric acid profiling in chrono-sampled artificial sweat (left) tracked by SERS intensity at 635 cm\u003csup\u003e-1\u003c/sup\u003e. Colored bars represent calibrated SERS intensity of input uric acid concentration. Optical images of the sequential sampling over time depending on the chamber volume (blue: 1.5 μL, red: 0.5 μL), which visualize sparse and dense sampling intervals (right). The scale bars represent 5 mm. Machine-learned quantification of (\u003cstrong\u003ec\u003c/strong\u003e) uric acid, (\u003cstrong\u003ed\u003c/strong\u003e) lactate, and (\u003cstrong\u003ee\u003c/strong\u003e) tyrosine in the mixture presented by prediction depending on true concentration. (center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range) Extracted SHAP feature importance (FI, solid line) and SERS spectra at 10 mM (dotted line) of (\u003cstrong\u003ef\u003c/strong\u003e) uric acid, (\u003cstrong\u003eg\u003c/strong\u003e) lactate, and (\u003cstrong\u003eh\u003c/strong\u003e) tyrosine. The bars represent the matched SERS peaks (colored) and notches (gray).\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/1036ae1aea64a810d64c680c.png"},{"id":76088296,"identity":"b022fb12-598c-4a7c-a458-a5bd60a73132","added_by":"auto","created_at":"2025-02-12 08:04:03","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":6564515,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLabel-free human sweat profiling of assorted metabolites.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Schematic illustration of on-body evaluation. Optical images of (\u003cstrong\u003eb\u003c/strong\u003e) on-body evaluation (left) and the CEP-SERS patch attached to human skin (right), (\u003cstrong\u003ec\u003c/strong\u003e) sequential sweat sampling (top), and sample preservation under different physical distortions (bottom). The scale bar represents 10 mm. Accuracy of label-free quantification for (\u003cstrong\u003ed\u003c/strong\u003e) uric acid and (\u003cstrong\u003ee\u003c/strong\u003e) lactate in the human sweat. (\u003cstrong\u003ef\u003c/strong\u003e) Comparison between tyrosine concentrations depending on participants measured by CMA (light gray) and SERS (dark gray). (\u003cstrong\u003eg-i\u003c/strong\u003e) Chronological profiling of each metabolite during exercise (left) measured by SERS (red rectangles) and FMA (blue circles) presented with schematic illustrations of each metabolic pathway (right). The error bars represent one standard deviation from the mean. (\u003cstrong\u003ej\u003c/strong\u003e) Measured heart rate (HR, black solid line), RER (orange bar: 0.7 \u0026lt; RER \u0026lt; 0.85), oxygen (VO\u003csub\u003e2\u003c/sub\u003e, blue dotted line), and carbon dioxide (VCO\u003csub\u003e2\u003c/sub\u003e, red dotted line) intake during on-body evaluation to monitor exercise intensity. Concentration comparison of (\u003cstrong\u003ek\u003c/strong\u003e) uric acid and (\u003cstrong\u003el\u003c/strong\u003e) tyrosine under different physiological conditions. (center line, median; box limits, upper and lower quartiles; whiskers, 1.5x interquartile range; cross, outlier; point, data) (\u003cstrong\u003em\u003c/strong\u003e) Relative metabolic phenotyping of individuals (I\u003csub\u003eN\u003c/sub\u003e, thin solid line) and averaged result (bold solid line) under fasting conditions (blue) and after purine-rich diet intake (red).\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/d5f62049583d412b50e7397c.png"},{"id":90072064,"identity":"2ca672ca-9dee-415b-afd6-420fb3e020ab","added_by":"auto","created_at":"2025-08-28 07:09:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":30525322,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/a2ecd01a-eec5-4b2d-bdb5-4b3011ccfb91.pdf"},{"id":76088308,"identity":"c4a3a596-499b-488f-8063-99f91e816db3","added_by":"auto","created_at":"2025-02-12 08:04:04","extension":"mp4","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":46501635,"visible":true,"origin":"","legend":"Supplementary video 1","description":"","filename":"SupplementaryVideo1.mp4","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/bcf57a48282f214fb62dfe51.mp4"},{"id":76088325,"identity":"8807dc61-f392-4156-9d01-2c701539fc9c","added_by":"auto","created_at":"2025-02-12 08:04:05","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":93824617,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary information updated\u003c/p\u003e","description":"","filename":"SupplementaryInformationKHJeongXKAIST.docx","url":"https://assets-eu.researchsquare.com/files/rs-5624954/v1/21e55f8070f7b828cda87db6.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"All-Flexible Chronoepifluidic Nanoplasmonic Patch Allows Label-free Sweat Profiling","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMetabolic phenotyping is pivotal in precision medicine, revealing individual health traits for personalized interventions [1]. Metabotypes surpass genetic predispositions, offering valuable insights into the diversity of individuals influenced by extrinsic determinants such as behavioral patterns, regimens, or gut microbial activity [2,3]. The interrelation of the metabolite reflects precision health status through metabolic pathway alterations induced by physiological changes or disorders. [4\u0026ndash;8]. In particular, transient metabolic alterations provide a physiological cue for proactive healthcare. Such cues facilitate the instant recognition of physiological disorders to promote optimal health outcomes. For instance, postprandial metabolic rates provide crucial information on individual responses to dietary intake, aiding in the development of personalized nutrition plans [9]. In addition, exercise or lifecycle routines can be tailored by monitoring daily activity-induced metabolite fluctuations [10]. However, conventional approaches profile static metabolite in blood [11,12] or urine [13\u0026ndash;15], which poses challenges in dynamic metabolic profiling.\u003c/p\u003e\n\u003cp\u003eRecent biosensing wearables profile the metabolic kinetics in biological fluids such as tear fluid [16,17], saliva [18], or sweat [19,20]. Sweat, unlike others, exhibits chemical abundance [21,22] and simple sample collection with less contamination [23], thus allowing in-situ profiling of metabolite alteration in proactive healthcare [24]. For example, analyzing lactate dynamics suggests optimal exercise routines by evaluating the lactate threshold or maximal lactate steady state during physical activities [25\u0026ndash;27]. In addition, postprandial changes of branched-chain amino acid or uric acid provide prognostic cues for metabolic syndrome [28\u0026ndash;30] or gout [31], respectively. Moreover, cortisol levels [32\u0026ndash;34] indicate psychological stress responses during daily life. Furthermore, sweat is easily collected by epidermal interfaces with simple absorbents [35\u0026ndash;37], or microfluidic patches [38\u0026ndash;40], exhibiting high compatibility with assorted wearable sensors. Such epidermal sweat sensors often utilize electrochemical [26,27,30,31,33,34,36,38,41] or colorimetric approaches [42\u0026ndash;45]. However, molecular recognition elements such as antibodies or enzymes on active sensing sites still hinder multiple and unveiled biomarkers detection for a comprehensive understanding of an individual\u0026apos;s unique physiological profile.\u003c/p\u003e\n\u003cp\u003eSurface-enhanced Raman spectroscopy (SERS) allows label-free and quantitative detections of diverse biochemicals [46\u0026ndash;48]. Plasmonic nanostructures are integrated into flexible epidermal patches for label-free metabolic profiling through sweat. [48,49,50,51] Furthermore, recent advancements in plasmonic patches combine with epidermal microfluidic channels for not only transient analysis [49\u0026ndash;57] but also in-situ collection and chronological profiling of metabolites [58,59,60]. For instance, rigid SERS substrates are frequently integrated into elastomeric microfluidic patches [58,61] to analyze metabolic alterations [58]. Partial rigidity of the SERS substrate still constrains conformal skin contact and mechanical durability in wearable applications, whereas plasmonic paper-based microfluidic sensors feature high flexibility thus overcoming such technical limitations. However, such sensors still lack delicate microfluidic control [48,52,55,62,63], sensitivity, and specificity [59,64], thus restricting the direct quantification of multiple analytes in complex sweat. In particular, such SERS patches encounter compatibility issues with sophisticated functional microfluidics such as capillary bursting valves [65\u0026ndash;67], facilitating sequential sampling for sweat profiling. As a result, the all-flexible epifluidic SERS patch still remains a technical challenge in precise and multiplexed sweat profiling.\u003c/p\u003e\n\u003cp\u003eHere we report an all-flexible chronoepifluidic SERS patch (CEP-SERS patch) for label-free profiling of sweat metabolites. The CEP-SERS patch consists of a plasmofluidic channel layer (PCL) for sweat collection, storage, and SERS analysis, and a dermal contact layer (DCL) for skin attachment (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). PCL includes a flexible plasmonic SERS substrate and a microfluidic sequential sampler. The SERS substrate features plasmonic nanoislands on ultrathin fluorocarbon-coated PDMS membrane, driven by low-temperature solid-state dewetting of thin silver film. The ultrathin fluorocarbon film effectively dewets the thin metal film on the sequential sampler, resulting in plasmonic structures with strong electromagnetic hotspots for highly sensitive SERS analysis. In addition, the sequential sampler serially collects the sweat via the microfluidic channel with capillary bursting valves, which spatially separate the sweat over time. Furthermore, DCL contains medical adhesive with a sweat collection port and interconnects the PCL on the skin for stable sweat collection. Finally, the CEP-SERS patch allows machine-learned label-free quantification of multiple metabolites in chrono-sampled sweat, thus quantitatively profiling sweat over time during physical activities. Such all-flexible feature ensures conformal contact for on-skin chronological sweat collection and label-free quantification of metabolites.\u003c/p\u003e\n\u003cp\u003eThe CEP-SERS patch was fabricated by integrating micro- and nanofabricated PCL and DCL (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). First, an ultrathin fluorocarbon layer was coated on the polydimethylsiloxane (PDMS) chronoepifluidic sweat sampler via atmospheric pressure chemical vapor deposition (AP-CVD) (S1). The fluorocarbon coating significantly reduces the surface energy and increases the surface roughness, thereby facilitating the low-temperature solid-state dewetting of an Ag thin film. Subsequently, a 10 nm-thick Ag thin film was thermally evaporated onto the fluorocarbon-coated sweat sampler (S2). The evaporated Ag thin film was thermally dewetted at 160\u0026deg;C for 30 minutes to form Ag nanoislands (S3). Steps S2 and S3 were repeated to enhance the plasmonic hotspots on the PCL. Finally, an air-venting outlet was punched on the PCL attached to the DCL with a sweat port for encapsulation and skin attachment (S4). Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ec and \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ed present optical images of the fabricated PCL, including an inset scanning electron microscopy (SEM) image of the Ag nanoislands, and the CEP-SERS patch on the skin. Details and stepwise optical images of the fabrication procedure are described in the Supplementary Materials (Fig. \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e"},{"header":"Result","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eAll-flexible SERS Substrate: Ag nanoislands on ultrathin fluorocarbon\u003c/h2\u003e\n \u003cp\u003eThe plasmonic properties of Ag nanoislands on the CEP-SERS patch are precisely controlled by the thicknesses of fluorocarbon and Ag thin film. Figure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea presents SEM images (left) and E-field distribution (middle) of Ag nanoislands, calculated by the finite-difference time-domain method (Lumerical FDTD Solutions) method. The structural features (right top) and absorption properties (right bottom) are shown for 10 nm Ag film on PDMS without fluorocarbon coating (green), single-dewetted 10 nm Ag film (red), and double-dewetted 10 nm Ag film (blue) on fluorocarbon coated PDMS. Unlike bare PDMS, the fluorocarbon intermediate layer provides low surface energy, leading to an enlargement of nanoisland diameters by facilitating lower-temperature solid-state dewetting of thin film (Fig. S2). In addition, two-step dewetting further increases the nanoisland diameters and the packing density up to 40%. These morphological modifications induce a redshift of plasmon resonance (Fig. S3) and enhance E-field intensity for highly-sensitive SERS detection.\u003c/p\u003e\n \u003cp\u003eThe fluorocarbon thickness is determined by signal-to-noise ratio (SNR) in SERS measurements. Figure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb shows the SERS intensity of 10 \u0026micro;M Rhodamine 6G (R6G) at 1365 cm⁻\u0026sup1; and SNR depending on the fluorocarbon thickness. A thick fluorocarbon layer enhances the SERS intensity (Fig. S4a) but introduces SERS background noise (Fig. S4b). Therefore, the fluorocarbon thickness was optimized to 2 nm to maximize the SNR, defined as the ratio of the SERS peak intensity at 1365 cm⁻\u0026sup1; to the standard deviation of the fluorocarbon SERS noise signals. Next, the Ag film thickness is optimized for the maximum SERS intensity. Figure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec shows SERS intensity for 1 \u0026micro;M R6G at 1365 cm⁻\u0026sup1; depending on the Ag film thickness and dewetting repetition. A thick Ag film facilitates the formation of large, densely packed nanoislands during the dewetting process (Fig. S5), leading to enhanced SERS intensity (blue bar). Repeated dewetting further enhances the SERS intensity (red bar) by providing abundant plasmonic hotspots. Note that the coalescence of nanoislands at a total thickness of 24 nm leads to a drastic decline in the SERS signal (Fig. S6). As a result, the CEP-SERS patch is fabricated by a 10 nm repeated dewetting process, which provides the maximum SERS intensity (Fig. S7). The SERS intensity of R6G exhibits strong linearity with concentration (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ed and Fig. S8). The SERS substrate provides an average SERS enhancement factor of 1.8 \u0026times; 10\u003csup\u003e7\u003c/sup\u003e with a uniformity of 11.8% (Fig. S9). The CEP-SERS patches are vacuum-sealed for further on-body analysis (Fig. S10) and over 85% of the performance is preserved for up to 25 days (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ee). The mechanical stability of the CEP-SERS patch is confirmed by measuring the SERS intensity of benzenethiol at 1070 cm⁻\u0026sup1;. The SERS performance remains stable with no notable degradation after 200 cycles of twisting (red) and bending (blue) (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ef). The CEP-SERS patch remains stable and effective for wearable sweat profiling.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eMicrofluidic Sequential Sampler\u003c/h3\u003e\n\u003cp\u003eThe sequential sweat sampler allows chronological sample collection through capillary bursting valves (CBVs) in a microfluidic network. Figure \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea illustrates the sampling sequence (left) and SEM images of the microfabricated CBVs (right). The bursting pressure (BP) gradients guide a sample fluid into different chambers for sequential sampling. The BPs are controlled by the width and diverging angle of the CBVs. In particular, the narrow width and the high diverging angle increase the BP in the hydrophobic channel (Fig. S11). The sample fluid enters the first chamber through CBV 1 and fills the chamber. Once the chamber is filled, the fluid then flows into the subsequent chamber through CBV 2. CBV 3 has the highest BP for air ventilation during the sampling. Note that air ventilation allows stable sweat collection without the formation of bubbles. The microfluidic parameters are determined by two-phase fluid dynamic analysis (COMSOL Multiphysics 6.1, Fig. S12). The sequential loading of distinct colored solutions visualizes the stepwise procedure of chrono-sample collection (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb). In addition, the sampling interval is determined by the chamber volume and the flow rate (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ec). Each chamber sequentially stores a collected sample in a separate space, and the small chamber volume allows dense sampling intervals over time (Fig. S13 and Supplementary Video 1). In this experiment, the sweat patch with a microchamber volume of 0.5 \u0026micro;L fulfills each chamber within ~\u0026thinsp;2 minutes at a flow rate of 0.25 \u0026micro;L/min.\u003c/p\u003e\n\u003cp\u003eThe collected samples are preserved without mixing through on-chip isolation and evaporation suppression. The isolation prevents diffusion between fluids, allowing for long-term storage without mixing (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ed). An air pocket barrier creates trapped microbubbles that separate each sample. The effectiveness of isolation is demonstrated by measuring SERS intensity from the deionized water (DI water) in the chamber after sequentially injecting DI water and R6G solution. The isolated chambers maintain initial concentration for 30 hours (red), while the absence of the air pocket barrier leads to mixing (blue). Low-permeable polyester film, attached to the medical adhesive at the sequential sampler, prevents evaporation. This method retains about 80% of samples after 30 hours. Note that evaporation-induced loss is quantified by measuring fluid and air bubble fractions (Fig. S14). This microfluidic sampler facilitates stable sweat collection on the skin and long-term preservation.\u003c/p\u003e\n\u003ch3\u003eMachine-learned label-free quantification of metabolites\u003c/h3\u003e\n\u003cp\u003eThe CEP-SERS patch allows label-free quantification of assorted metabolites. Figure \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ea shows the SERS peak intensities of uric acid (red, at 635 cm\u003csup\u003e-1\u003c/sup\u003e), lactate (blue, at 859 cm\u003csup\u003e-1\u003c/sup\u003e), and tyrosine (green, at 1353 cm\u003csup\u003e-1\u003c/sup\u003e) depending on the different concentrations in the range of physiological level (SERS spectra are presented in Fig. S15). A strong linear relationship between SERS peak intensity and concentration facilitates label-free metabolite quantification. Note that intrinsic Raman background noises from PDMS were initially removed for accurate signal interpretation of metabolites. The metabolic profiling over time is further shown by tracking the changes in the SERS peak intensity of the target molecules within artificial sweat samples (Fig. S16a). The uric acid concentration is sequentially adjusted to 80 \u0026micro;M, 10 \u0026micro;M, and 20 \u0026micro;M over the injection time by using a syringe pump (Fig. S16b). Figure \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eb illustrates the variation in SERS peak intensity of uric acid in chrono-sampled artificial sweat over the collection period (left), utilizing the CEP-SERS patch with chamber volumes of 1.5 \u0026micro;L (blue) and 0.5 \u0026micro;L (red) (SERS spectra are presented in Fig. S16c and d). The CEP-SERS patch with a small chamber volume effectively improves the precise quantification of rapid variations in target molecular concentration by providing dense sampling intervals (right).\u003c/p\u003e\n\u003cp\u003eMachine-learned label-free quantification of metabolites is further conducted by using an autoencoder with a logistic regression model. The autoencoder-based prediction model is trained to minimize the total loss function, including the reconstruction and prediction losses (Fig. S17) [68]. Target metabolites including uric acid, lactate, and tyrosine are mixed with 41 different combinations of concentration, and a total of 1,476 spectra are provided for robust machine-learned quantification, which considers various background states of sweat (Fig. S18 and S19). The SERS spectra mapped onto the two-dimensional (2D) latent space via autoencoder show that the individual classes are diagonally aligned depending on the concentration of each metabolite (Fig. S20). Figure \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ec-e shows the predicted concentrations and the corresponding true values for uric acid (R\u003csup\u003e2\u003c/sup\u003e: 0.71\u0026ndash;0.80), lactate (R\u003csup\u003e2\u003c/sup\u003e: 0.65\u0026ndash;0.83), and tyrosine (R\u003csup\u003e2\u003c/sup\u003e: 0.82\u0026ndash;0.92), which are evaluated with 10 times of repeated random sampling cross-validation (Fig. S21). The concentration of each metabolite is well predicted to the true value and the quantification process is explained through the feature extraction calculated by the Shapley additive explanation (SHAP) value (Fig. S22) [69,70]. The SHAP feature importance (solid line) is presented with corresponding SERS spectra at 10 mM concentration (dotted line) for uric acid (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003ef), lactate (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eg), and tyrosine (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eh). The extracted features contain the matched characteristics to the SERS spectra including SERS peaks (colored bar) and notches (gray bar). This validation demonstrates the machine-learned quantifications are explainable and established based on the molecule-specific SERS signals.\u003c/p\u003e\n\u003ch3\u003eLabel-free human sweat profiling of assorted metabolites\u003c/h3\u003e\n\u003cp\u003eThe CEP-SERS patch collects human sweat on the skin and label-freely profiles the transient alteration of metabolites. The on-body evaluation includes a treadmill warm-up followed by a climb mill exercise, performed on separate days under fasting conditions and after a purine-rich diet intake (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ea, Fig. S23). The CEP-SERS patches are attached to multiple sites on the forehead and shoulder of participants\u0026apos; skin to collect exercise-induced sweat (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eb). The sweat samples are collected during the evaluation through the CEP-SERS patch (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ec). The inset images show the sweat collection using the patch without plasmonic structures to clearly present the sequential sampling. All flexibility of the patch ensures conformal dermal contact, resulting in stable sequential sweat collection (top). In addition, the CEP-SERS patch securely preserves sweat samples under various physical stresses that may occur in a wearable environment such as compressing, twisting, or detaching (bottom). Note that the biophysical signals such as heart rate, respiratory exchange ratio (RER), oxygen, and carbon dioxide intake are simultaneously measured during the evaluation to monitor the exercise intensity.\u003c/p\u003e\n\u003cp\u003eThe CEP-SERS patch performs label-free profiling of chrono-sampled human sweat using machine-learned quantification models for individual metabolites. Sweat samples from four healthy participants are analyzed to assess metabolic differences during exercise with a purine-rich meal versus fasting. The average SERS signals measured from human sweat capture key features of various metabolites in sweat (Fig. S24). SERS signals from chrono-sampled sweat are obtained seven times from each chamber to reduce measurement errors (Fig. S25 - S28). The sweat flow rate for each participant is determined by dividing the total volume of the collected samples by the sampling time (Fig. S29). Label-free uric acid, lactate, and tyrosine quantifications in human sweat are validated using commercial fluorometric assay (FMA) or colorimetric assay (CMA) kits. The sweat is further collected by swiping the microtube to the forehead for FMA and CMA. Machine-learned predictions for uric acid and lactate demonstrate strong agreement with FMA measurements, achieving R\u0026sup2; values of 0.96 and 0.86, respectively (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ed and e). In addition, tyrosine prediction from participant 2\u0026ndash;4 shows an error margin of ~\u0026thinsp;5 \u0026micro;M to the CMA kit (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ef). For Participant 1, contaminants including skin lipids during sample collection significantly affected the absorbance properties, resulting in a substantial difference of ~\u0026thinsp;25 \u0026micro;M compared to the CMA measurement. The metabolites in the collected sweat samples (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003eg-i) are chronologically profiled with biophysical signals (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ej) through the machine-learned SERS quantification models (Chronological profiling results of metabolites and biophysical signals for four participants under fasting conditions and purine-rich diet intake are presented in Fig. S30 - S33). The CEP-SERS patch captures the metabolic alterations associated with the metabolic pathway of each metabolite and dilution. The metabolites in sweat are often diluted over time during perspiration [24,31,34,71]. In contrast, lactate temporarily increases, which reflects the eccrine sweat gland metabolism [24,26,72] and anaerobic metabolism [73,74] during exercise. The dietary intake increases the overall concentrations of uric acid [31] and tyrosine (amino acid) [30,71,75] in the on-body evaluation (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003ek and \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003el). The metabolic alterations under fasting (blue) and purine-rich diet intake conditions (red) show physiological states for four participants. The averaged alteration (bold red line) reveals that lactate levels remain stable due to minimal dietary influence. In contrast, uric acid and tyrosine increase post-ingestion, reflecting physiological changes in purine metabolism and protein digestion, respectively (Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003em). The CEP-SERS patch combined with machine-learned metabolic quantification has successfully captured transient physiological changes by label-free and multiplexed detection of exercise- and intake-induced metabolic alterations.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eMetabolic profiling offers significant insights into individual physiological states, enhancing the scope of digital phenotyping and advancing personalized healthcare. Wearable sweat sensors non-invasively capture the transient biochemical alterations during assorted activities and allow on-body metabolic profiling. Furthermore, SERS holds significant potential to unveil comprehensive physiological information by facilitating label-free universal molecular recognition in sweat. Unlike others (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), our CEP-SERS patch features the integration of nanoplasmonics and functional microfluidics for precise chronological profiling of metabolites. The plasmonic nanostructures are integrated into the microfluidic sampler via large-area fabrication of Ag nanoislands on fluorocarbon-coated PDMS, facilitating both SERS analysis and stable sequential sweat collection. In addition, SERS spectra from chrono-sampled sweat samples are profiled through a robust machine-learned quantification model that accounts for various concentration combinations of the background in sweat. The CEP-SERS patch has successfully captured the exercise- and intake-induced alterations in multiple metabolites. This CEP-SERS patch combined with machine-learned quantification can provide a new strategy for delineating the physiological phenotype of individuals in personalized healthcare.\u003c/p\u003e "},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eNumerical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eE-field distribution.\u0026nbsp;\u003c/strong\u003eThe electric fields of Ag nanoislands with varying geometries were numerically calculated utilizing a three-dimensional finite-difference time-domain method (Lumerical FDTD Solutions). Geometric parameters of the Ag nanoislands were derived from binary segmented SEM images employing Image J software, with Ag nanoislands modeled as a cylindrical shape.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFluid dynamics.\u0026nbsp;\u003c/strong\u003eFluid dynamics at the capillary bursting valve were numerically calculated by utilizing finite elements methods (FEM, \u0026nbsp; COMSOL Multiphysics 6.1). The calculations were conducted by integrating the two-phase flow and level-set physics modules. In addition, the two-phase flow, level set, and wetted wall phenomena were coupled to derive comprehensive computational results. A time-dependent solver was utilized to compute the temporal evolution of the sequential sampling.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMaterials and reagents\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRhodamine 6G (R6G, R4127-25G, dye content ~95%), benzenethiol (W361607-SAMPLE-K, \u0026ge;98%), sodium L-lactate (L7022-5G, ~98%), urea (U5378-100G), tyrosine (PHR1097-1G, certified reference material), glycine (G7126-100G, \u0026ge;99%), L-alanine (A7469-25G, \u0026ge;98.5%), L-glutamic acid monosodium salt monohydrate (49621-250G, \u0026ge;98.0%), uric acid (U2625-25G, \u0026ge;99%), L-ascorbic acid (A5960-25G, \u0026ge;99.0%), D-(+)-glucose (G8270-100G, \u0026ge;99.5%),\u0026nbsp;creatinine (PHR1462-1G, certified reference material), sodium chloride (S7653-250G,\u0026nbsp;\u0026ge;99.5%), and potassium chloride (P5405-250G,\u0026nbsp;\u0026ge;\u0026nbsp;99.0%) are purchased from Sigma Aldrich. Medical adhesives were purchased from 3M (Tegaderm 1622W, 1522).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFabrication of CEP-SERS patch\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA Si wafer was immersed in buffered oxide etchant for 30 seconds to remove the native oxide layer. Photoresist (SU-8 2000.5, MicroChem Corp.) was then spin-coated to a thickness of 500 nm to form an adhesion layer. An additional photoresist (SU-8 2100, MicroChem Corp.) was applied to fabricate a microfluidic channel mold with a thickness of 200 \u0026micro;m. A 10:1 mixture of PDMS base and curing agent (Sylgard 184, Dow Corning Corp.) was spin-coated onto the microfluidic channel mold at 100 rpm for 30 seconds and then at 160 rpm for 30 seconds. The PDMS was cured on a hotplate at 80\u0026deg;C for 1 hour and 30 minutes. The PDMS layer was peeled off from the mold after curing and washed in isopropyl alcohol for 5 minutes. The fluorocarbon layer was coated on the microfluidic channel via AP-CVD with parameters set to 150 W plasma power, a helium flow rate of 5.0 L/min, and a fluorocarbon flow rate of 2 sccm (IHP-1000, APP Korea). Ag film was thermally deposited at a deposition rate of 1.0 \u0026Aring;/s by using a thermal evaporator (SNTEC Inc., Korea) and dewetted on a hotplate at 160\u0026deg;C for 30 minutes. The Ag deposition and dewetting steps were repeated twice to create strong plasmonic hotspots. The air ventilation outlet and sweat inlet were punched on PCL and medical adhesive film (Tegaderm 1622W, 3M) for DCL, respectively. The two layers were bonded by double-sided medical adhesive (1522, 3M) to complete the\u0026nbsp;CEP-SERS\u0026nbsp;patch fabrication.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCharacterization and measurement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAbsorption measurement.\u0026nbsp;\u003c/strong\u003eAbsorption spectra were measured by using a microscopic spectrometer setup comprising an inverted microscope (Axiovert 200M, Carl Zeiss ) integrated with a white light LED lamp (MCWHL5-C4, Thorlabs Inc.) and a spectrometer (MicroSpec 2300i) featuring a charge-coupled device (CCD) camera (Model PIXIS: 400BR, Princeton Instruments). Absorption spectra of Ag nanoislands on fluorocarbon-coated PDMS were acquired utilizing a 20\u0026times; objective lens (NA=0.5).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSERS measurement.\u0026nbsp;\u003c/strong\u003eA helium-neon laser (HRP050, Thorlabs Inc.) operating at a wavelength of 633\u0026thinsp;nm was utilized in conjunction with a spectrometer featuring a CCD camera, both integrated with an inverted microscope. Light excitation and collection were performed via a 20\u0026times; objective lens (NA\u0026thinsp;= 0.5). The excitation laser was operated at a power of 5\u0026thinsp;mW, while data acquisition times were set to 1\u0026thinsp;second for characterization and extended to 10 seconds for metabolites and human sweat sample measurements. Note that a\u0026nbsp;single SERS spectrum from human sweat was acquired by averaging 6 times of the measurement to minimize signal variances. All the SERS spectra are measured as solution state, and an equivalent volume of DI water was applied to the sample before detection for dry samples.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine-learned metabolic quantification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSERS data preparation.\u0026nbsp;\u003c/strong\u003eFirst, 1,476 spectra with 41 different concentration combinations including target molecules are provided for machine-learned quantification. The measured SERS spectra have a vector length of 1321, which contains the signal intensities for the Raman shift range of 457\u0026ndash;1674 cm\u003csup\u003e\u0026ndash;1\u003c/sup\u003e. Each vector was normalized to a range of 0-1.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMetabolic quantification.\u0026nbsp;\u003c/strong\u003eThe model consists of a symmetric encoder and decoder with four equally sized layers of 1321 nodes, maintaining a consistent number of neurons across all layers, with a latent layer of length 2 between the encoder and decoder.\u0026nbsp;The model was trained with the Adam optimizer, and the hyperparameters were heuristically tuned to optimize performance. Initially, the learning rate, weight decay, and batch size were set to 1e\u003csup\u003e-4\u003c/sup\u003e, 1e\u003csup\u003e-5\u003c/sup\u003e, and 32 respectively. The maximum number of training epochs was set to 150, with early stopping employed. After the 50\u003csup\u003eth\u003c/sup\u003e epoch, a scheduled weight decay adjustment was applied to promote further model generalization, reducing the learning rate and weight decay to 1e\u003csup\u003e-5\u003c/sup\u003e and 1e\u003csup\u003e-6\u003c/sup\u003e, respectively. The model predicts the concentration of each metabolite based on the values along the concentration axis in the latent space.\u0026nbsp;Note that three-quarters of the entire SERS spectra data were used for training, with the remainder reserved for validation. In addition, 10 times of repeated random sampling cross-validation ensures robust model evaluation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSERS feature importance calculation.\u0026nbsp;\u003c/strong\u003eSHAP (SHapley Additive exPlanations) values were utilized to identify the significant features for concentration prediction. The SHAP calculations were performed using only the encoder part of the model, focusing on the feature importance independent of the increase or decrease in concentration. For each sample, the SHAP values were squared and then averaged across all samples to obtain the overall importance of each feature, thereby identifying key spectral features contributing to the model\u0026apos;s predictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOn-body evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtocols.\u003c/strong\u003e The participant performed the following exercise protocol on separate days under fasting conditions and after the purine-rich diet intake, which included 125 g to 250 g of sardines. On the days when the purine-rich diet was consumed, the exercise was initiated after a 1.5-hour rest period. The participants wear the CEP-SERS patches attached to the forehead and shoulder, a wearable metabolic system (Cosmed K5, Cosmed) for respiratory gas analysis, and a heart rate sensor (Polar OH1, Polar Electro Oy) for heart rate monitoring. The participant performed a warm-up consisting of two cycles of running on a treadmill at a 7.2 km/hr speed for 4 minutes, followed by a 3-minute rest period. Subsequently, the participant engaged in exercise on a climbmill with progressively increasing intensity until the heart rate reached 80% of the maximum heart rate. The participant completed an additional 5-minute rest period after exercise, concluding the protocol. Note that sweat samples were additionally collected using a microtube swapped on the forehead to measure the actual concentration of sweat using the fluorometric assay or colorimetric assay kit. Sweat was collected once after the treadmill exercise, at 3-5 minute intervals during the climb mill exercise, and once after a 5-minute rest period following the completion of all exercise tasks. All human trials were approved by the Institutional Review Board of Korea Advanced Science and Technology (protocol number: KH2024-085). In addition, we obtained informed consent from all participants before the experiment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFluorometric assays for validation of uric acid and lactate quantification.\u003c/strong\u003e The concentrations of uric acid and lactate in collected sweat were validated by using a commercial uric acid assay kit (MAK077-1KT, Sigma Aldrich) and a lactate assay kit (BM-LAC-100, PicoSens\u003csup\u003eTM\u003c/sup\u003e). For uric acid measurement, 5 \u0026micro;L of sweat sample was mixed with 45 \u0026micro;L of uric acid assay buffer, followed by the addition of 2.17 \u0026micro;L of probe and 2.17 \u0026micro;L of enzyme mix. The mixture was thoroughly mixed by pipetting and incubated for 30 minutes at 37\u0026deg;C. For lactate measurement, sweat samples were diluted 200-fold with deionized water. A 10 \u0026micro;L aliquot of the diluted sweat sample was mixed with 40 \u0026micro;L of lactate assay buffer, followed by the addition of 0.43 \u0026micro;L of probe and 2.17 \u0026micro;L of enzyme mix. The mixture was thoroughly mixed by pipetting and incubated for 30 minutes at room temperature. Fluorescence intensity was measured at excitation and emission wavelengths of 535 nm and 590 nm, respectively. Uric acid and lactate concentrations were calculated using a standard curve of each molecule.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eColorimetric assays for validation of tyrosine quantification.\u003c/strong\u003e The concentrations of tyrosine in collected sweat were validated by using a commercial tyrosine assay kit (MET-5073, Cell Biolabs Inc.). Sweat samples were prepared by adjusting their volume to 50 \u0026micro;L with standard tyrosine solutions. The mixed samples were chilled on ice for 30 minutes and then centrifuged at 10,000 g for 15 minutes at 4\u0026deg;C. The supernatant was carefully transferred to prevent the pellet from dissolving and subsequently passed through a QIAquick Spin Column (28115) by centrifugation (MICRO 17 TR, Hanil Science Co.) at 5,000 g for 10 minutes at 4 \u0026deg;C. The filtered samples were then used for tyrosine quantification. The 10x enzyme from the kit was diluted to the assay solution, and 15 \u0026micro;L of each prepared sample and standard tyrosine solution was added to individual wells of a transparent 384-well plate, followed by 15 \u0026micro;L of the diluted enzyme solution to make a total volume of 30 \u0026micro;L per well. The plate was incubated at room temperature for 10 minutes on a horizontal shaker, and absorbance was measured at 490 nm using a plate reader (CYTATION 5 imaging reader, Agilent BioTek). Tyrosine concentrations were then calculated using a standard curve.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eDisclosures \u003c/strong\u003eThe authors declare no conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u0026nbsp;The authors declare that the main data supporting the findings of this study are available within the article and its Supplementary Information file. The accession codes for raw SERS spectra will be available before publication.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability Code\u003c/strong\u003e\u0026nbsp;underlying the results presented in this paper is not publicly available at this time but may be obtained from the authors upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments.\u003c/strong\u003e This work was supported by the National Research Foundation of Korea (NRF) funded by the Ministry of Science ICT \u0026amp; Future Planning (2022M3H4A4085645, RS-2024-00438316), the Technology Innovation Program (No. RS-2024-00432381) funded by the Ministry of Trade Industry \u0026amp; Energy (MOTIE, Korea).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eAuthors and Affiliations\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eDepartment of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea\u003c/p\u003e\n\u003cp\u003eJaehun Jeon, \u0026nbsp;Sangyeon Lee, Hanjin Kim, Eun-Sil Yu, Hamin Na, Doheon Lee \u0026amp; Ki-Hun Jeong\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eKAIST Institute for Health Science and Technology (KIHST), KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon, Republic of Korea, 305-701\u003c/p\u003e\n\u003cp\u003eJaehun Jeon, Eun-Sil Yu, Hamin Na \u0026amp; Ki-Hun Jeong\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDepartment of Mechanical Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon, 34141, Korea\u003c/p\u003e\n\u003cp\u003eSeongok Chae \u0026amp; Hyung-Soon Park\u003c/p\u003e\n\u003cp\u003eBionanotechnology Research Center, Korea Research Institute of Bioscience and Biotechnology (KRIBB), 125 Gwahak-ro, Yuseong-gu, Daejeon 34141, Republic of Korea\u003c/p\u003e\n\u003cp\u003eJoo Hoon Lee \u0026amp; Taejoon Kang\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eContributions\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eJ.J. and K.-H.J. conceived the idea and designed the projects. K.-H.J. supervised the overall research. J.J. designed, fabricated, characterized, and packaged the all-flexible chronoepifluidic SERS patch and conducted overall experiments. S.L., H.K., and D.L. contributed to the machine-learned quantification of metabolites. J.J., S.C., and H.-S.P. conducted the on-body evaluation of the SERS patch. J.J., J.H.L., and T.K. validated the SERS patch using the commercial fluorometric and colorimetric kit. J.J., K.-H.J., S.L., S.C., J.H.L., H.K., E.-S.Y., H.N., T.K., H.-S.P., and D.L. evaluated the experiments and contributed valuable ideas. J.J. and K.-H.J. wrote the manuscript. All authors discussed the results and commented on the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eCorresponding author\u003c/strong\u003e\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eCorrespondence to Ki-Hun Jeong.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eHolmes, E., Wilson, I. D., and Nicholson, J. K. Metabolic phenotyping in health and disease. Cell, 134(5), 714\u0026ndash;717 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolmes, E., Loo, R. L., Stamler, J., Bictash, M., Yap, I. K., Chan, Q., \u0026hellip; and Elliott, P. Human metabolic phenotype diversity and its association with diet and blood pressure. Nature, 453(7193), 396\u0026ndash;400 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNicholson, J. K., Holmes, E., Kinross, J. M., Darzi, A. W., Takats, Z., and Lindon, J. C. Metabolic phenotyping in clinical and surgical environments. Nature, 491(7424), 384\u0026ndash;392 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumas, M. E., Kinross, J., and Nicholson, J. K. Metabolic phenotyping and systems biology approaches to understanding metabolic syndrome and fatty liver disease. Gastroenterology, 146(1), 46\u0026ndash;62 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolmes, E., Wijeyesekera, A., Taylor-Robinson, S. D., and Nicholson, J. K. The promise of metabolic phenotyping in gastroenterology and hepatology. Nature reviews Gastroenterology \u0026amp; hepatology, 12(8), 458\u0026ndash;471 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEverett, J. R., Holmes, E., Veselkov, K. A., Lindon, J. C., and Nicholson, J. K. A unified conceptual framework for metabolic phenotyping in diagnosis and prognosis. Trends in Pharmacological Sciences, 40(10), 763\u0026ndash;773 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuhre, K. Metabolic profiling in diabetes. J Endocrinol, 221(3), R75-R85 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eInoue, K., Tsutsui, H., Akatsu, H., Hashizume, Y., Matsukawa, N., Yamamoto, T., and Toyo'Oka, T. Metabolic profiling of Alzheimer's disease brains. Scientific reports, 3(1), 2364 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSecor, S. M. Specific dynamic action: a review of the postprandial metabolic response. Journal of Comparative Physiology B, 179, 1\u0026ndash;56 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEmerson, S. R., Kurti, S. P., Emerson, E. M., Cull, B. J., Casey, K., Haub, M. D., and Rosenkranz, S. K. Postprandial metabolic responses differ by age group and physical activity level. The Journal of nutrition, health and aging, 22(1), 145\u0026ndash;153 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFiehn, O. and Kind, T. Metabolite profiling in blood plasma. Metabolomics: Methods and protocols, 3\u0026ndash;17 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrauslys, A., Phelan, M. M., Broughton, C., Baines, P. B., Jennings, R., Siner, S., \u0026hellip; and Carrol, E. D. NMR-based metabolic profiling provides diagnostic and prognostic information in critically ill children with suspected infection. Scientific Reports, 10(1), 20198 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWant, E., Wilson, I., Gika, H. et al. Global metabolic profiling procedures for urine using UPLC\u0026ndash;MS. Nature Protocols 5, 1005\u0026ndash;1018 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMath\u0026eacute;, E. A., Patterson, A. D., Haznadar, M., Manna, S. K., Krausz, K. W., Bowman, E. D., \u0026hellip; and Harris, C. C. Noninvasive urinary metabolomic profiling identifies diagnostic and prognostic markers in lung cancer. Cancer research, 74(12), 3259\u0026ndash;3270 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuan, H., Liu, L. F., Tang, Z., Zhang, M., Chua, K. K., Song, J. X., \u0026hellip; and Cai, Z. Comprehensive urinary metabolomic profiling and identification of potential noninvasive marker for idiopathic Parkinson\u0026rsquo;s disease. Scientific reports, 5(1), 13888 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark, W., Seo, H., Kim, J. et al. In-depth correlation analysis between tear glucose and blood glucose using a wireless smart contact lens. Nature Communications 15, 2828 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeum, D. H., Kim, S. K., Koo, J., Lee, G. H., Jeon, C., Mok, J. W., \u0026hellip; and Hahn, S. K. Wireless smart contact lens for diabetic diagnosis and therapy. Science Advances, 6(17), eaba3252 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, J., Imani, S., de Araujo, W. R., Warchall, J., Vald\u0026eacute;s-Ram\u0026iacute;rez, G., Paix\u0026atilde;o, T. R., \u0026hellip; and Wang, J. Wearable salivary uric acid mouthguard biosensor with integrated wireless electronics. Biosensors and Bioelectronics, 74, 1061\u0026ndash;1068 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrunmair, J., Gotsmy, M., Niederstaetter, L., Neuditschko, B., Bileck, A., Slany, A., \u0026hellip; and Gerner, C. Finger sweat analysis enables short interval metabolic biomonitoring in humans. Nature Communications, 12(1), 5993 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBariya, M., Li, L., Ghattamaneni, R., Ahn, C. H., Nyein, H. Y. Y., Tai, L. C., and Javey, A. Glove-based sensors for multimodal monitoring of natural sweat. Science advances, 6(35), eabb8308 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTu, J., Torrente-Rodr\u0026iacute;guez, R. M., Wang, M. and Gao, W. The era of digital health: a review of portable and wearable affinity biosensors. Advanced Functional Materials, 30, 1906713 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeikenfeld, J. et al. Accessing analytes in biofluids for peripheral biochemical monitoring. Nature Biotechnology, 37, 407\u0026ndash;419 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eD. S. Yang, R. Ghaffari, and J. A. Rogers Sweat as a diagnostic biofluid. Science, 379(6634), 760\u0026ndash;761 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMin, J., Tu, J., Xu, C., Lukas, H., Shin, S., Yang, Y., \u0026hellip; and Gao, W. Skin-interfaced wearable sweat sensors for precision medicine. Chemical reviews, 123(8), 5049\u0026ndash;5138 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, G., Hong, J., and Park, S. B. Wearable device for continuous sweat lactate monitoring in sports: a narrative review. Frontiers in Physiology, 15, 1376801 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKomkova, M. A., Eliseev, A. A., Poyarkov, A. A., Daboss, E. V., Evdokimov, P. V., Eliseev, A. A., \u0026amp; Karyakin, A. A. Simultaneous monitoring of sweat lactate content and sweat secretion rate by wearable remote biosensors. Biosensors and Bioelectronics, 202, 113970 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXuan, X., Chen, C., Molinero-Fernandez, A., Ekelund, E., Cardinale, D., Swar\u0026eacute;n, M., \u0026hellip; and Crespo, G. A. Fully integrated wearable device for continuous sweat lactate monitoring in sports. ACS sensors, 8(6), 2401\u0026ndash;2409 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLynch, C., Adams, S. Branched-chain amino acids in metabolic signaling and insulin resistance. Nature Reviews Endocrinology 10, 723\u0026ndash;736 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayers, J., Wu, C., Clish, C. et al. Elevation of circulating branched-chain amino acids is an early event in human pancreatic adenocarcinoma development. Nature Medicine 20, 1193\u0026ndash;1198 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eM. Wang, Y. Yang, J. Min, Y. Song, J. Tu, D. Mukasa, \u0026hellip; and W. Gao A wearable electrochemical biosensor for the monitoring of metabolites and nutrients. Nature Biomedical Engineering, 6(11), 1225\u0026ndash;1235 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, Y., Song, Y., Bo, X. et al. A laser-engraved wearable sensor for sensitive detection of uric acid and tyrosine in sweat. Nature Biotechnology 38, 217\u0026ndash;224 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRussell, E., Koren, G., Rieder, M., and Van Uum, S. H. The detection of cortisol in human sweat: implications for measurement of cortisol in hair. Therapeutic drug monitoring, 36(1), 30\u0026ndash;34 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParlak, O., Keene, S. T., Marais, A., Curto, V. F., and Salleo, A. Molecularly selective nanoporous membrane-based wearable organic electrochemical device for noninvasive cortisol sensing. Science advances, 4(7), eaar2904 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, W., Yin, L., Sempionatto, J. R., Moon, J. M., Teymourian, H., and Wang, J. Touch-based stressless cortisol sensing. Advanced Materials, 33(18), 2008465 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eParrilla, M., Guinovart, T., Ferr\u0026eacute;, J., Blondeau, P., and Andrade, F. J. A wearable paper-based sweat sensor for human perspiration monitoring. Advanced Healthcare Materials, 8(16), 1900342 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNagamine, K., Mano, T., Nomura, A., Ichimura, Y., Izawa, R., Furusawa, H., \u0026hellip; and Tokito, S. Noninvasive sweat-lactate biosensor emplsoying a hydrogel-based touch pad. Scientific reports, 9(1), 10102 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao, F. J., Bonmarin, M., Chen, Z. C., Larson, M., Fay, D., Runnoe, D., and Heikenfeld, J. Ultra-simple wearable local sweat volume monitoring patch based on swellable hydrogels. Lab on a Chip, 20(1), 168\u0026ndash;174 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNyein, H.Y.Y., Bariya, M., Tran, B. et al. A wearable patch for continuous analysis of thermoregulatory sweat at rest. Nature Communications 12, 1823 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReeder, J.T., Xue, Y., Franklin, D. et al. Resettable skin interfaced microfluidic sweat collection devices with chemesthetic hydration feedback. Nature Communications, 10, 5513 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReeder, J. T., Choi, J., Xue, Y., Gutruf, P., Hanson, J., Liu, M., \u0026hellip; and Rogers, J. A. Waterproof, electronics-enabled, epidermal microfluidic devices for sweat collection, biomarker analysis, and thermography in aquatic settings. Science Advances, 5(1), eaau6356 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao, W., Emaminejad, S., Nyein, H. et al. Fully integrated wearable sensor arrays for multiplexed in situ perspiration analysis. Nature 529, 509\u0026ndash;514 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eS. Emaminejad, W. Gao, W. Wu, Z. A. Davies, H. Yin Yin Nyein, S. Challa, \u0026hellip; and R. W. Davis Autonomous sweat extraction and analysis applied to cystic fibrosis and glucose monitoring using a fully integrated wearable platform. \u003cem\u003eProceedings of the National Academy of sciences\u003c/em\u003e, 114(18), 4625\u0026ndash;4630 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR. T. Ray, M. Ivanovic, P. M. Curtis, D. Franklin, K. Guventurk, W. J. Jeang, \u0026hellip; and J. A. Rogers Soft, skin-interfaced sweat stickers for cystic fibrosis diagnosis and management. Science translational medicine, 13(587), eabd8109 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoh, A., Kang, D., Xue, Y., Lee, S., Pielak, R. M., Kim, J., \u0026hellip; and Rogers, J. A. A soft, wearable microfluidic device for the capture, storage, and colorimetric sensing of sweat. Science translational medicine, 8(366), 366ra165-366ra165 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, J., Wu, Y., Luan, H., Yang, D. S., Cho, D., Kwak, S. S., \u0026hellip; and Rogers, J. A. A skin-interfaced, miniaturized microfluidic analysis and delivery system for colorimetric measurements of nutrients in sweat and supply of vitamins through the skin. Advanced Science, 9(2), 2103331 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, Y., Zhao, C., Wang, J., Luo, X., Xie, L., Zhan, S., \u0026hellip; and Ying, Y. Wearable plasmonic-metasurface sensor for noninvasive and universal molecular fingerprint detection on biointerfaces. Science Advances, 7(4), eabe4553 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBi, X., Czajkowsky, D.M., Shao, Z. et al. Digital colloid-enhanced Raman spectroscopy by single-molecule counting. Nature (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, W., Li, Q., Ren, J., Feng, K., Gong, J., Li, Z., \u0026hellip; and Yang, L. A sensor platform based on SERS detection/janus textile for sweat glucose and lactate analysis toward portable monitoring of wellness status. Biosensors and Bioelectronics, 263, 116612 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, H., Cui, Q., Xu, L., Tian, Y., Jiao, A., Wang, C., \u0026hellip; and Chen, M. Silk fibroin fibers decorated with urchin-like Au/Ag nanoalloys: a flexible hygroscopic SERS sensor for monitoring of folic acid in human sweat. Optics Express, 29(19), 30892\u0026ndash;30904 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, X., Wang, X., Ning, M., Wang, P., Wang, W., Zhang, X., \u0026hellip; and Li, S. Fast synthesis of Au nanoparticles on metal\u0026ndash;phenolic network for sweat SERS analysis. Nanomaterials, 12(17), 2977 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoh, E. H., Lee, W. C., Choi, Y. J., Moon, J. I., Jang, J., Park, S. G., \u0026hellip; and Jung, H. S. A wearable surface-enhanced Raman scattering sensor for label-free molecular detection. ACS Applied Materials and Interfaces, 13(2), 3024\u0026ndash;3032 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, Z., Wang, W., Tian, H., Yu, W., Niu, Y., Zheng, X., \u0026hellip; and Huang, Y. Wearable intelligent sweat platform for SERS-AI diagnosis of gout. Lab on a Chip, 24(7), 1996\u0026ndash;2004 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYe, H., Chen, X., Huang, X., Li, C., Yin, X., Zhao, W., and Wang, T. Patterned Gold Nanoparticle Superlattice Film for Wearable Sweat Sensors. Nano Letters, \u003cem\u003e24\u003c/em\u003e(35), 11082\u0026ndash;11089 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiong, S., Wang, C., Zhu, C., Dong, P., and Wu, X. Dual Detection of Urea and Glucose in Sweat Using a Portable Microfluidic SERS Sensor with Silver Nano-Tripods and 1D-CNN Model Analysis. ACS Applied Materials \u0026amp; Interfaces (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, M., Zhu, K., Wei, J., Yang, K., Wu, L., Zong, S., and Wang, Z. Silk fibroin-based wearable SERS biosensor for simultaneous sweat monitoring of creatinine and uric acid. Biosensors and Bioelectronics, \u003cem\u003e265\u003c/em\u003e, 116662 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuan, P. C., Qi, Q. J., Wang, Y. Q., Lin, J. S., Zhang, Y. J., and Li, J. F. Development of a 3D Hydrogel SERS Chip for Noninvasive, Real-Time pH and Glucose Monitoring in Sweat. ACS Applied Materials \u0026amp; Interfaces, \u003cem\u003e16\u003c/em\u003e(36), 48139\u0026ndash;48146 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, H., Zhang, H., Sikdar, D., Liu, X., Yang, Z., Cheng, W., and Chen, Y. Jellyfish-like Gold Nanowires as FlexoSERS Sensors for Sweat Analysis. Nano Letters, \u003cem\u003e24\u003c/em\u003e(36), 11269\u0026ndash;11278 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe, X., Fan, C., Luo, Y., Xu, T., and Zhang, X. Flexible microfluidic nanoplasmonic sensors for refreshable and portable recognition of sweat biochemical fingerprint. npj Flexible Electronics, 6(1), 60 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMogera, U., Guo, H., Namkoong, M., Rahman, M. S., Nguyen, T., and Tian, L. Wearable plasmonic paper\u0026ndash;based microfluidics for continuous sweat analysis. Science advances, 8(12), eabn1736 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, G., Zhao, X., Tang, X., Yao, L., Li, W., Wang, J.,and Hao, Q. Wearable Hydrogel SERS Chip Utilizing Plasmonic Trimers for Uric Acid Analysis in Sweat. Nano Letters, \u003cem\u003e24\u003c/em\u003e(42), 13447\u0026ndash;13454 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiao, J., Wang, J., Luo, Y., Xu, T., and Zhang, X. \u003cem\u003eWearable plasmonic sweat biosensor for acetaminophen drug monitoring\u003c/em\u003e. ACS sensors, 8(4), 1766\u0026ndash;1773 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJain, V., Ochoa, M., Jiang, H. et al. A mass-customizable dermal patch with discrete colorimetric indicators for personalized sweat rate quantification. Microsystems and Nanoengineering 5, 29 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng, Y., Feng, S., Ning, Q. et al. Dual-signal readout paper-based wearable biosensor with a 3D origami structure for multiplexed analyte detection in sweat. Microsystems and Nanoengineering 9, 36 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y., Guo, Y., Chen, H., Xiao, X., Long, F., Zhong, H., \u0026hellip; and Liu, Z. Flexible Wearable Plasmonic Paper-Based Microfluidics with Expandable Channel and Adjustable Flow Rate for Portable Surface-Enhanced Raman Scattering Sweat Sensing. ACS Photonics (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi, J., Kang, D., Han, S., Kim, S. B., and Rogers, J. A. Thin, soft, skin-mounted microfluidic networks with capillary burst valves for chrono‐sampling of sweat. Advanced healthcare materials, 6(5), 1601355 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, C. H., Ma, H. J. H., Baessler, P., Balanay, R. K., and Ray, T. R. Skin-interfaced microfluidic systems with spatially engineered 3D fluidics for sweat capture and analysis. Science Advances, 9(18), eadg4272 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, Y., Kanatzidis, E. E., Avila, R., Zhou, M., Bai, Y., Chen, S., \u0026hellip; and Rogers, J. A. 3D-printed epidermal sweat microfluidic systems with integrated microcuvettes for precise spectroscopic and fluorometric biochemical assays. Materials Horizons, 10(11), 4992\u0026ndash;5003 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, S., Kim, H., and Lee, D. Linearization autoencoder: an autoencoder-based regression model with latent space linearization. bioRxiv, 2022\u0026ndash;06 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg, S. M., Erion, G., Chen, H., DeGrave, A., Prutkin, J. M., Nair, B., \u0026hellip; and Lee, S. I. From local explanations to global understanding with explainable AI for trees. Nature machine intelligence, 2(1), 56\u0026ndash;67 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, H., Covert, I. C., Lundberg, S. M., \u0026amp; Lee, S. I. Algorithms to estimate Shapley value feature attributions. Nature Machine Intelligence, 5(6), 590\u0026ndash;601 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhong, B., Qin, X., Xu, H., Liu, L., Li, L., Li, Z., \u0026hellip; and Wang, L. Interindividual-and blood-correlated sweat phenylalanine multimodal analytical biochips for tracking exercise metabolism. Nature Communications, 15(1), 624 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDerbyshire, P. J., Barr, H., Davis, F., and Higson, S. P. Lactate in human sweat: a critical review of research to the present day. The journal of physiological sciences, 62, 429\u0026ndash;440 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXuan, X., Perez-Rafols, C., Chen, C., Cuartero, M., and Crespo, G. A. Lactate biosensing for reliable on-body sweat analysis. ACS sensors, 6(7), 2763\u0026ndash;2771 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarpova, E. V., Laptev, A. I., Andreev, E. A., Karyakina, E. E., and Karyakin, A. A. Relationship between sweat and blood lactate levels during exhaustive physical exercise. ChemElectroChem, 7(1), 191\u0026ndash;194 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFernstrom, M. H. and Fernstrom, J. D. Protein consumption increases tyrosine concentration and in vivo tyrosine hydroxylation rate in the light-adapted rat retina. Brain research, 401(2), 392\u0026ndash;396 (1987).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":false,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5624954/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5624954/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWearable sweat sensors allow non-invasive metabolic profiling for timely intervention in proactive healthcare. However, molecular recognition elements in conventional sensors still hinder a comprehensive understanding of an individual's unique physiological profile. Here we report an all-flexible chronoepifluidic surface-enhanced Raman spectroscopy (SERS) patch (CEP-SERS patch) for label-free sweat profiling. The CEP-SERS patch features the integration of nanoplasmonics and functional microfluidics for precise chronological profiling of metabolites. An ultrathin fluorocarbon film facilitates large-area nanofabrication of plasmonic structures on a functional microfluidic channel via low-temperature solid-state dewetting of a thin silver film. The CEP-SERS patch facilitates conformal contact on human skin and SERS detection of diverse metabolites from sequentially sampled sweat. Machine-learned quantification of metabolites including lactate, uric acid, and tyrosine has successfully profiled SERS detection of sweat during assorted physical activities. This CEP-SERS patch can provide a new strategy for delineating the physiological phenotype of individuals in personalized healthcare.\u003c/p\u003e","manuscriptTitle":"All-Flexible Chronoepifluidic Nanoplasmonic Patch Allows Label-free Sweat Profiling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-12 08:03:58","doi":"10.21203/rs.3.rs-5624954/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"07ac56b4-e41c-42ef-91b0-6d42bcaee2a6","owner":[],"postedDate":"February 12th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":44152572,"name":"Biological sciences/Biotechnology/Nanobiotechnology/Nanofabrication and nanopatterning"},{"id":44152573,"name":"Physical sciences/Nanoscience and technology/Nanobiotechnology/Microfluidics"},{"id":44152574,"name":"Physical sciences/Engineering/Biomedical engineering"},{"id":44152575,"name":"Physical sciences/Materials science/Materials for optics/Nanophotonics and plasmonics"}],"tags":[],"updatedAt":"2025-08-28T07:08:44+00:00","versionOfRecord":{"articleIdentity":"rs-5624954","link":"https://doi.org/10.1038/s41467-025-63510-2","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2025-08-27 04:00:00","publishedOnDateReadable":"August 27th, 2025"},"versionCreatedAt":"2025-02-12 08:03:58","video":"","vorDoi":"10.1038/s41467-025-63510-2","vorDoiUrl":"https://doi.org/10.1038/s41467-025-63510-2","workflowStages":[]},"version":"v1","identity":"rs-5624954","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5624954","identity":"rs-5624954","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}