Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath

arXiv:2604.05707 · physics.med-ph, cs.LG · Submitted 2026-04-07 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath".

Title and authors: Kai: We discussed how this paper on "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath" goes beyond just using acetone by screening for other VOCs that could act as markers for fat oxidation.

Mira: Right, the core summary is that they performed an untargeted screening using PTR-TOF-MS during exercise and after a ninety minute recovery period to look for volatile organic compounds besides acetone.

Lev: So they were using this mass spectrometry setup to analyze what was exhaled during the cycling sessions and then again after a short rest period.

Kai: They found seven hundred seventy three relevant signals in that analysis, which is the raw output from their automated signal detection algorithm.

Mira: But they only found four of those signals that showed a strong correlation with blood BOHB levels, which they used as their reference marker for fat oxidation.

Lev: That means out of all those potential markers, only four actually matched up well with the known biological process they were tracking.

Kai: Those four signals had a rho value greater than zero point eight two and a p value of zero point zero zero zero two, which is how they established statistical significance for those specific measurements.

Mira: And what was interesting was that visual inspection helped them figure out that the six other weakly correlated signals were just signal artifacts, which means they filtered out some of the noise.

Lev: So the methodology involved taking a lot of raw data and running it through an automated process to isolate the relevant peaks before looking for correlations.

Kai: And those four strongly correlated signals turned out to be all attributable to acetone or its isotopologues or fragments, which confirms acetone’s central role in this finding.

Mira: So the summary is that while they searched broadly, the actual biological signal they tied back to was primarily related to acetone levels.

Lev: It means the real discovery isn't necessarily a brand new molecule, but confirming and strengthening what we already knew about acetone’s utility in this context.

The paper's summary: Kai: The paper points out a few ways they improved their approach, starting with the signal detection algorithms used to process the PTR-TOF-MS data.

Mira: They integrated automated signal detection algorithms directly into the data pre processing step to automatically detect relevant peak signals and map them to mass values, which improves the efficiency of untargeted analysis.

Lev: Which is smart because manually sorting seven hundred seventy three signals would take forever; you need that kind of automation to make this feasible in a lab setting.

Kai: And then they moved on to the predictive modeling part, where they used end-of-exercise measurements of these specific acetone signals to train K-nearest neighbour regression models.

Mira: That prediction model takes those end of exercise values and uses them as training data to predict the follow up blood BOHB levels, which is the core mechanism for making a forward looking statement.

Lev: So the improvement isn't just finding new things; it’s building a system that can actually use the data they already have to make a prediction about future outcomes.

Kai: They also created classification systems using F1 scores and accuracy based on the prediction that participants would exhibit substantial post exercise changes in BOHB, which helps identify people who are likely to change significantly.

Mira: That classification system gives you a concrete metric, an F1 score of zero point eight three and an accuracy of zero point eight nine for identifying those with a BOHB change greater than one point two.

Lev: That F1 score is the metric we need to look at when we think about translating this into something that runs on actual hardware; it tells us how good the classifier is at separating those positive cases from the negative ones.

Kai: So they’ve improved by moving from just observation to creating a system that can make a prediction about future BOHB changes based on during exercise measurements.

The paper's improvements: Mira: So to wrap up, the main conclusion of this paper on "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath" is that basic predictions of eventual fat oxidation may be feasible during training based solely on BrAce measurements.

Kai: It means we can make some basic statements about the expected fat oxidation already during exercising, and that's based only on the BrAce measurements they collected.

Lev: From an engineering view, it suggests that this approach is viable for developing monitoring devices that could provide users with real time projections of anticipated fat oxidation during physical activity.

Mira: It confirms acetone as the key marker and shows we can get some basic predictions about what’s going to happen with fat oxidation later, which is a step toward anticipating physiological changes instead of just reacting to them.

Lev: I think we should keep thinking about how much error we can tolerate before this becomes something that's usable in a field, because accuracy matters.

Mira: It’s important to remember they didn't find any novel breath based biomarkers of fat oxidation, meaning they didn't discover anything entirely new outside of what was expected.

Kai: So the paper "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath" shows that basic predictions of eventual fat oxidation may be feasible during training based solely on BrAce measurements.

Conclusion: Kai: So, to wrap up, this paper "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath" confirms that basic predictions of eventual fat oxidation may be feasible during training based solely on the BrAce measurements they collected.

Mira: Right, it suggests we can make some basic statements about what’s going to happen with fat oxidation already during exercising, and that's based only on the BrAce measurements. It's about using this breath data to predict future blood BOHB changes.

Lev: From an engineering view, it suggests that this approach is viable for developing monitoring devices that could provide users with real time projections of anticipated fat oxidation during physical activity. But we gotta remember the fidelity of those predictions matters a lot on real hardware.

Kai: Yeah, I mean they trained their K-nearest neighbour regression models on end of exercise signal values to predict follow up BOHB levels, and they got an F1 score of zero point eight three and an accuracy of zero point eight nine when identifying participants with substantial post exercise BOHB changes.

Mira: That F1 score is the metric we need to look at when we think about translating this into something that runs on actual hardware. It shows how good the classifier is at separating those positive cases from the negative ones.

Lev: And they also developed a classification system based on that prediction, which helps identify people who are likely to change significantly in their BOHB levels after exercise. That's a useful tool for screening.

Kai: It’s important to remember they didn't find any novel breath based biomarkers of fat oxidation, meaning they didn't discover anything entirely new outside of what was expected from acetone and related compounds.

Mira: True, the main finding was confirming acetone as the key marker and showing that other VOCs could be screened, but only four signals actually correlated strongly with BOHB.

Lev: So the methodology involved taking a lot of raw data and running it through an automated process to isolate the relevant peaks before looking for correlations with blood BOHB concentrations. That’s how they got those numbers.

Kai: I think that’s what makes this interesting, though, because it moves beyond just using acetone as a standalone marker during exercise and tries to find other things that might be more specific.

Mira: It does, but the study didn't reveal any novel breath based biomarkers of fat oxidation, so the focus is really on improving what we already knew about acetone’s utility in this context.

Lev: So it’s a solid piece of work for setting up a baseline prediction framework for fatigue markers during exercise. We can definitely take that framework and start figuring out how to build some actual sensors that can actually run those K-nearest neighbour models reliably.

Kai: Exactly, so the paper "Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath" gives us a starting point for what might be possible with wearable monitoring.

Mira: Next up, we've got something looking at gradient based inverse design of free energy landscapes with diffusion models. That’s a different kind of predictive modeling, but it’s also trying to build those kinds of complex models to understand physical systems better.

André Homeyer, Júlia Blanka Sziládi, Jan-Philipp Redlich, Jonathan Beauchamp, Y Lan Pham

Fraunhofer Institute for Digital Medicine MEVIS · Department of Sensory Analytics and Technologies, Fraunhofer Institute for Process Engineering and Packaging IVV

physics.med-ph, cs.LG

Submitted: 2026-04-07

Updated: 2026-04-07

DOI: 10.1088/1752-7163/aec0c3

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 75/100

The gist: Breath acetone represents a promising non-invasive biomarker for monitoring fat oxidation during exercise The study performed an untargeted screening for volatile organic compounds (VOCs) that could…

Key concepts

Fat Oxidation
This refers to the process where the body burns fat for energy during physical activity. It is measured using ketone bodies like acetoacetate, β-hydroxybutyrate (BOHB), and acetone, which are produced in the liver when carbohydrate availability is low.
Volatile Organic Compounds (VOCs)
These are gases that easily evaporate and can be detected in breath. The study screened many different VOCs using mass spectrometry to see if any of them could act as biomarkers for fat burning, similar to how acetone works.
BrAce
This term refers to the measurements taken from exhaled breath during exercise, primarily focusing on acetone. The study explored whether these specific breath signals can be used not just as a measure of current fat burning but also to predict future changes in fat oxidation levels.

Terminology

Summary

Breath acetone represents a promising non-invasive biomarker for monitoring fat oxidation during exercise The study performed an untargeted screening for volatile organic compounds (VOCs) that could serve as markers of fat oxidation beyond acetone and investigated whether breath measurements taken during exercise could predict post-exercise changes in fat oxidation

Study Design and Methodology

The researchers performed an untargeted screening for VOCs using proton-transfer-reaction time-of-flight mass spectrometry (PTR-TOF-MS) during exercise and after a 90-min recovery period Nineteen participants completed two 25-min cycling sessions separated by a brief 5-min rest period Blood β-hydroxybutyrate (BOHB) concentrations served as the reference marker for fat oxidation Breath was sampled using a buffered end-tidal (BET) breath sampler connected to the PTR-TOF-MS instrument Blood BOHB concentrations were determined by taking capillary blood samples from the fingertips of participants using FORA Sterile Lancets and applying them to FORA β-Ketone test strips

Signal Detection and Identification

The automated signal detection algorithm identified 773 relevant signals in PTR-TOF-MS measurements The analysis focused on correlations between the relative changes of the relevant signals and BOHB at the follow-up time point Only four signals (m/z 31.027, 59.046, 59.214 and 60. p. 5

Where the paper says it

Only four signals (m/z 31. 214 and 60.

051) showed strong correlations with BOHB levels (ρ ≥ 0. p. 5
Where the paper says it

214 and 60. 051) showed strong correlations with BOHB levels (ρ ≥ 0.

82, p = 0.0002) Visual scrutiny of the data allowed the six weakly correlated signals to be attributed to signal artefacts The four strongly correlated signals could all be linked to acetone, with two signals attributed to protonated acetone and its 13C-isotopologue, respectively

Predictive Modeling of Fat Oxidation

The study assessed whether post-exercise changes in blood BOHB reference measurements could already be predicted from breath measurements during exercise For the two acetone signals (m/z 59.046 and m/z 60. p. 5

Where the paper says it

Evidently, the signals occur at different magnitudes, with m/z 59. 046 being by far the most abundant.

051), end-of-exercise values were used to train K-nearest neighbour regression models to predict follow-up BOHB levels Using end-of-exercise values of protonated acetone, participants with substantial post-exercise changes in BOHB (≥1. p. 5
Where the paper says it

To assess the predictability of follow-up BOHB levels from end-of-exercise signal values, K-nearest neighbour regression models were trained and applied using leave-one-out cross-validation. For each signal and each participant, a distinct model was trained on data from all other participants, using the end-of-exercise signal values as input and the corresponding follow-up BOHB levels as output.

2) could be correctly identified with an F1 score of 0. p. 6
Where the paper says it

Using end-of-exercise values of protonated acetone, participants with substantial post-exercise changes in BOHB (≥1. 2) could be correctly identified with an F1 score of 0.

83 (accuracy: 0. p. 0
Where the paper says it

End-of-exercise measurements of these signals enabled accurate prediction of participants with substantial post-exercise BOHB changes (F1 score ≥ 0. 83, accuracy = 0.

89)

Conclusion and Future Implications

The findings suggest that basic predictions of eventual fat oxidation may be feasible during training, based solely on BrAce measurements The study suggests that it could be possible to make basic statements about the expected fat oxidation already during exercising and solely based on BrAce measurements Future research should investigate more thoroughly whether BrAce measurements during exercise can serve as reliable predictors of post-exercise fat oxidation If subsequent research validates these findings more broadly, this could enable the development of innovative BrAce monitoring devices that provide users with real-time projections of anticipated fat oxidation during physical activity The authors concluded that basic predictions of eventual fat oxidation may be feasible during training, based solely on BrAce measurements

--- Page 1 ---

Untargeted analysis of volatile markers of post-exercise fat oxidation in exhaled breath

André Homeyer1

, Júlia Blanka Sziládi2

, Jan-Philipp Redlich1

, Jonathan Beauchamp2

, Y Lan Pham2,∗

1Fraunhofer Institute for Digital Medicine MEVIS, Bremen, Germany 2Department of Sensory Analytics and Technologies and Technologies IVV Department of Sensory Analytics and Technologies and Technologies Freising Germany Abstract Breath acetone represents a promising non-invasive biomarker for monitoring fat oxidation during exercise However, its utility is limited by confounding factors, as well as by the fact that significant changes in concentration occur only hours post-exercise, which makes real-time assessment difficult We performed an untargeted screening for volatile organic compounds (VOCs) that could serve as markers of fat oxidation beyond acetone and investigated whether breath measurements taken during exercise could predict post-exercise changes in fat oxidation Nineteen participants completed two 25-min cycling sessions separated by a brief 5-min rest period VOC emissions were analysed using proton-transfer-reaction time-of-flight mass spectrometry (PTR-TOF-MS) during exercise and after a 90 min recovery period Blood β-hydroxybutyrate (BOHB) concentrations served as the reference marker for fat oxidation Among 773 relevant analytical features detected in the PTR-TOF-MS measurements, only four signals exhibited strong correlations with BOHB (ρ ≥ 0.82, p = 0.0002)—all attributable to acetone or its isotopologues or fragments End-of-exercise measurements of these signals enabled accurate prediction of participants with substantial post-exercise BOHB changes (F1 score ≥ 0. p. 0

Where the paper says it

0002)—all attributable to acetone or its isotopologues or fragments. End-of-exercise measurements of these signals enabled accurate prediction of participants with substantial post-exercise BOHB changes (F1 score ≥ 0.

83, accuracy = 0. p. 0
Where the paper says it

End-of-exercise measurements of these signals enabled accurate prediction of participants with substantial post-exercise BOHB changes (F1 score ≥ 0. 83, accuracy = 0.

89) Our study did not reveal any novel breath-based biomarkers of fat oxidation, but it confirmed acetone as the key marker Moreover, our findings suggest that breath acetone measurements during exercise may already enable basic predictions of post-exercise fat oxidation Keywords: acetone, fat burning, ketosis, PTR-TOF-MS, prediction, untargeted analysis

--- Page 2 ---

  1. Introduction

Overweight and obesity are associated with more than 1.2 million deaths annually in Europe and surrounding regions; on average, 63% of men and 54% of women in the EU are overweight, and the trend is rising [1] Physical activity is one of the cornerstones in reducing overweight and obesity [2] A variety of factors influence fat burning, including training intensity and duration, diet, fitness level and time of day [3–5] Therefore, it would be desirable to receive feedback on fat burning status during training in order to monitor exercise efficacy Fat burning, technically referred to as fat oxidation, is commonly quantified based on three ketone bodies: acetoacetate, β-hydroxybutyrate (BOHB) and acetone [8, 9] These molecules are produced in the liver as metabolites of free fatty acids When carbohydrate availability is reduced, for instance during fasting or prolonged exercise, systemic ketone body concentrations increase [9] and are therefore exploitable as biomarkers of fat oxidation [10] Acetone is the only ketone body that is volatile and can diffuse freely from the blood into the airways and thereby manifest in exhaled breath This makes acetone an attractive target for non-invasive monitoring of fat oxidation via breath Multiple studies have demonstrated strong, non-linear correlations between breath acetone (BrAce) concentrations and BOHB levels [8, 11, 13] One challenge that could limit the usefulness of BrAce as a biomarker for fat oxidation during exercise is its variability [17, 18] Consequently, it seems reasonable to explore exhaled breath for other VOCs that more specifically reflect fat oxidation

--- Page 3 ---

Participants 19

Sex [f/m] 6/13 Age [years] 30.2 ± 7.2 BMI [kg/m2] 23. p. 1

Where the paper says it

Consumption of alcohol or certain drugs, such as disulfiram, can increase BrAce 10- to 15-fold [18]. Personal characteristics, including body-mass index (BMI), sex, and exercise habits, as well as disease states, such as diabetes and infections, further contribute to BrAce variation [8, 18, 22, 23].

6 ± 2.7 Height [cm] 175. p. 2
Where the paper says it

2 BMI [kg/m2 ] 23. 7 Height [cm] 175.

5 ± 8.2 Weight [kg] 73. p. 2
Where the paper says it

7 Height [cm] 175. 2 Weight [kg] 73.

3 ± 13.1 Table 1: Cohort demographics and metadata Values are presented as mean ± standard deviation BMI = body mass index

--- Page 4 ---

Figure 1: Experimental protocol Left: Participants completed two 25-min cycling sessions on a cycle ergometer C1 Upright Lifecycle Life Fitness Europe GmbH Unterschleißheim Germany Centre: Exhaled breath was sampled via a buffered end-tidal (BET) sampler Right: PTR-TOF-MS measurements of exhaled breath were recorded throughout both cycling sessions and at follow-up 90 min post-exercise Blood BOHB concentrations were determined immediately before and after the cycling sessions and at follow-up

--- Page 5 ---

Figure 2: Data pre-processing The raw PTR-TOF-MS measurements of every participant were condensed into eleven 5 min steps For each step, a TOF spectrum was computed by masking the cycles belonging to end-tidal phases and averaging the respective TOF bin values The final harmonized data formed a three dimensional matrix in the format of 19 participants × 11 steps × 260,000 TOF bins with the values reflecting the ion counts recorded in each step and TOF bin

--- Page 6 ---

Figure 3: Distributions of the logarithmic intensities of the four significant signals at the follow-up measurement Data comprise the mean end-tidal plateau signal intensities of four exhalations at follow-up for each participant

--- Page 7 ---

Figure 4: Signal changes over time Top: Distributions of relative changes in signal intensity at m/z 59.046 (left) and m/z 60. p. 7

Where the paper says it

Top: Distributions of relative changes in signal intensity at m/z 59. 046 (left) and m/z 60.

051 (right) compared to the start of exercise Bottom: Correlation plots of the relative changes in m/z 59. p. 6
Where the paper says it

The distributions at follow-up were positively skewed, meaning most participants showed small increases, while a few had much larger ones. Figure 4, bottom row, plots the relative changes in protonated acetone compared to BOHB at the end of exercise (52.

046 and BOHB at the end of training (52. p. 7
Where the paper says it

Bottom: Correlation plots of the relative changes in m/z 59. 046 and BOHB at the end of training (52.

5 min; left) and the follow-up time point (147. p. 7
Where the paper says it

046 and BOHB at the end of training (52. 5 min; left) and the follow-up time point (147.

5 min; right)

--- Page 8 ---

Figure 5: Post-exercise predictability Top: Spearman correlations between relative changes in signal intensity of the two acetone signals (m/z 59.046, m/z 60. p. 5

Where the paper says it

Evidently, the signals occur at different magnitudes, with m/z 59. 046 being by far the most abundant.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Development of an untargeted analytical approach to screen for VOCs that could be potential markers of fat oxidation besides acetone, enabling discovery of novel breath-based biomarkers.

  2. Integration of signal detection algorithms into PTR-TOF-MS data pre-processing to automatically detect relevant peak signals and map them to mass values, improving the efficiency of untargeted analysis.

  3. Implementation of K-nearest neighbour regression models trained on end-of-exercise signal values to predict post-exercise blood BOHB levels, allowing for real-time estimation of anticipated fat oxidation during exercise.

  4. Creation of classification systems using F1 scores and accuracy based on the prediction that participants would exhibit substantial post-exercise changes in BOHB, enabling early identification of individuals with significant fat oxidation potential.

  5. Development of a framework that suggests basic predictions of eventual fat oxidation may be feasible during training, based solely on BrAce measurements, allowing users to receive real-time projections instead of waiting hours for post-exercise results.

Abstract

Breath acetone represents a promising non-invasive biomarker for monitoring fat oxidation during exercise. However, its utility is limited by confounding factors, as well as by the fact that significant changes in concentration occur only hours post-exercise, which makes real-time assessment difficult. We performed an untargeted screening for volatile organic compounds (VOCs) that could serve as markers of fat oxidation beyond acetone, and investigated whether breath measurements taken during exercise could predict post-exercise changes in fat oxidation. Nineteen participants completed two 25-min cycling sessions separated by a brief 5-min rest period. VOC emissions were analysed using proton-transfer-reaction time-of-flight mass spectrometry (PTR-TOF-MS) during exercise and after a 90-min recovery period. Blood β-hydroxybutyrate (BOHB) concentrations served as the reference marker for fat oxidation. Among 773 relevant analytical features detected in the PTR-TOF-MS measurements, only four signals exhibited strong correlations with BOHB (ρ at least 0.82, p = 0.0002)-all attributable to acetone or its isotopologues or fragments. End-of-exercise measurements of these signals enabled accurate prediction of participants with substantial post-exercise BOHB changes (F1 score at least 0.83, accuracy = 0.89). Our study did not reveal any novel breath-based biomarkers of fat oxidation, but it confirmed acetone as the key marker. Moreover, our findings suggest that breath acetone measurements during exercise may already enable basic predictions of post-exercise fat oxidation.

Related papers