Summary
Climbing plants use circumnutation, a rhythmic exploratory movement of the shoot apex, to locate supports.1–5 Although neighboring plants can alter this movement before physical contact, the underlying mechanisms remain unknown.6–9 Here, we combined three-dimensional kinematic tracking and real-time volatilomic profiling in Pisum sativum to investigate the temporal association between volatile emissions and circumnutation. Seven volatile-associated m/z features showed transient increases before peak circumnutation velocity in plants growing with a conspecific, a pattern absent in isolated plants. Neighbor presence also increased apex velocity, acceleration, path length, and explored area. Wavelet coherence analysis revealed stronger temporal coordination between real neighboring pairs than between randomly paired individuals. Neighboring plants additionally exhibited faster shoot growth and increased chlorophyll content in upper leaves. Together, these findings link volatile emission dynamics with exploratory movement and reveal coordinated behavioral and physiological responses to neighboring plants. This work establishes a framework for investigating the temporal organization of plant-plant interactions while highlighting the need for experiments to determine whether volatile cues causally influence movement.
Results and Discussion
Neighbor presence intensifies circumnutation and alters its temporal dynamics
To determine how neighboring plants influence exploratory growth, we compared circumnutation kinematics in Pisum sativum plants grown individually or alongside a conspecific. Plants grown with a neighbor exhibited significantly higher daily growth rates (estimate = +2.04 mm day−1; p = 0.00043), whereas overall growth showed limited temporal variation throughout the experiment. Generalized additive mixed models (GAMMs) revealed nonlinear temporal dynamics in all measured apex parameters (Table 1), indicating that movement features varied over time rather than following a constant rate of change.
Table 1. Summary of GAMM analyses of apex circumnutation parameters under individual and neighbor conditions.
| Kinematic parameter | Individual mean ± SD | Neighbor-condition mean ± SD | Global neighbor effect; does neighbor-mediated growth change the magnitude of the parameter? | Time dynamics; does this parameter change nonlinearly over time? | Dynamics differ between conditions; is the temporal trajectory altered by neighbor-mediated growth? |
|---|---|---|---|---|---|
| Velocity (apex) | 0.148 ± 0.032 | 0.222 ± 0.043 | neighbor > individual (β = 0.070, p < 0.001) | individual nonlinear (edf = 8.528, p < 0.001), neighbor nonlinear (edf = 8.930, p < 0.001) | yes (LRT=197.4, p < 0.001) |
| Acceleration max (apex) | 0.028 ± 0.004 | 0.038 ± 0.010 | neighbor > individual (β = 0.009, p = 0.001) | individual nonlinear (edf = 7.450, p < 0.001), neighbor nonlinear (edf = 8.010, p < 0.001) | no (LRT = 3.599, p = 0.166) |
| Acceleration min (apex) | –0.026 ± 0.004 | –0.034 ± 0.009 | individual > neighbor (β = –0.008, p = 0.004) | individual nonlinear (edf = 7.035, p < 0.001), neighbor nonlinear (edf = 7.629, p < 0.001) | yes (LRT = 10.05, p = 0.007) |
| Length (apex) | 16.849 ± 3.784 |
24.508 ± 5.068 | neighbor > individual (β = 7.28, p < 0.001) |
individual nonlinear (edf = 8.137, p < 0.001), neighbor nonlinear (edf = 8.806, p < 0.001) | yes (LRT = 47.72, p < 0.001) |
| Area (apex) | 28.490 ± 12.699 |
57.838 ± 21.364 | neighbor > individual (β = 29.958, p < 0.001) |
individual nonlinear (edf = 7.095, p < 0.001), neighbor nonlinear (edf = 8.894, p < 0.001) | yes (LRT = 79.59, p < 0.001) |
Descriptive statistics (mean ± SD) were calculated from plant-level averages for each condition. “Global neighbor effect” reports the parametric coefficient for growth condition from the GAMM and indicates the overall difference in parameter magnitude between the neighbor and individual conditions. “Time dynamics” describes the nonlinear temporal trajectories estimated using condition-specific smooth functions of time; edf indicates the estimated degrees of freedom and reflects the complexity of each temporal pattern. “Dynamics differ between conditions” reports the likelihood-ratio test (LRT) comparing models with shared and condition-specific smooths of time, and therefore tests whether the temporal trajectories differ between growth conditions.
These dynamics may reflect the combined influence of growth-related processes, endogenous oscillations, and environmental inputs. A representative example of this nonlinear trend is shown in Figure 1 for the circumnutation velocity of the apex.
Figure 1. Neighbor presence increases circumnutation velocity in climbing plants.

Raw measurements of apex circumnutation velocity are shown as semitransparent points. Solid and dashed lines represent generalized additive mixed model (GAMM) fits for the individual and neighbor conditions, respectively. Shaded bands indicate 95% CI. Time is expressed in minutes from the start of the experiment.
Neighbor presence significantly increased apex circumnutation velocity, maximum acceleration, length, and area (Table 1). These changes indicate that cues associated with neighbor presence enhance the intensity and spatial extent of exploratory movement. Increased velocity suggests that plants may accelerate support-search behavior when competitors are present, whereas greater circumnutation length and area indicate broader spatial exploration. Circumnutation parameters exhibited marked nonlinear temporal structure, highlighting the intrinsically dynamic nature of exploratory movement. Several apex-related parameters also displayed condition-specific temporal patterns, indicating that neighbor presence modifies both the magnitude and temporal structure of movement. Together, these results show that the presence of a conspecific competitor increases the exploratory capacity of climbing plants under competitive conditions.
Neighbor presence is associated with transient pre-peak increases in volatile m/z features
Comparative volatilomic analysis revealed marked differences between individual and neighbor-mediated growth conditions. Plants grown in the presence of a neighbor emitted a richer volatilome than isolated plants, with 100 m/z features showing mean intensities significantly greater than zero compared with 16 m/z features detected in individual plants (false discovery rate [FDR]-corrected tests; Table S1).
This difference indicates a broader set of detectable volatile features in the neighbor condition, although it should not be interpreted as direct evidence of increased emission rates or greater chemical diversity. Because chamber volume was identical across conditions, the presence of two plants in the neighbor condition may have increased overall volatile organic compound (VOC) concentrations and thereby enhanced the detectability of low-abundance m/z features.
We next investigated whether volatile dynamics were temporally associated with circumnutation. Seven m/z features showed significantly higher normalized intensities during the phase preceding the peak in apex circumnutation velocity than during both the baseline and peak phases (Figure 2A; Tables S2 and S3). These transient increases were observed in plants grown with a neighbor, whereas no features met the same criteria in individually grown plants. The seven m/z features were putatively assigned to benzenoids, organic oxygen compounds, or organoheterocyclic compounds. Because these assignments were based on direct-injection mass spectrometry, which lacks chromatographic separation, they should be considered tentative rather than confirmed chemical identifications. The time series of the seven m/z features are shown together with apex circumnutation velocity on a common time axis in Figure 2B.
Figure 2. Transient pre-peak increases in volatile emission and their temporal relationship with circumnutation velocity.

(A) Transient increases in m/z feature intensity before the peak in circumnutation velocity. Paired trajectories show the normalized intensity of each of the seven m/ z features across the baseline, pre-peak, and peak phases in plants grown with a neighbor. Each panel represents one feature, identified by its putative elemental formula and measured m/z value (Table S3). Colored lines connect phase-specific mean intensities for individual biological replicates (n = 14). Each biological replicate corresponds to a single growth chamber containing a dyad of two plants sharing a pot. Black circles indicate the mean across replicates, and error bars represent 95% CI. The trajectories illustrate the increase in feature intensity from baseline to pre-peak, followed by a decrease during the peak phase. Intensities were normalized to the benzene reagent-ion (C6H6+) signal and corrected by subtracting the volatile intensity measured in the blank chamber at each time point. Values are dimensionless normalized intensities expressed in arbitrary units. Pre-peak versus baseline and pre-peak versus peak comparisons were performed using paired one-sided t tests across replicates, with FDR correction across all detected m/z features (α = 0.05). Only features significant in both contrasts are shown (see also Tables S1, S2, S3, and S4).
(B) Temporal profiles of volatile emission and apex circumnutation velocity in plants grown with a neighbor. For each of the seven m/z features shown in (A), mean normalized feature intensity across replicates (red, left axis) is plotted alongside mean apex circumnutation velocity for the same replicates (dark blue, right axis) as a function of time from the start of the experiment. Shaded bands indicate 95% CI across replicates. Both time series are displayed as centered 6-h moving averages. The shaded vertical region indicates the pre-peak phase (days 4–6).
To verify that these dynamics were not affected by the light-dark cycle, we tested for day-night differences and repeated the phase comparisons after standardizing phase means for photoperiod. 6 of the 7 m/z features showed no significant day-night difference, and all 7 retained significant pre-peak increases in both comparisons after photoperiod standardization (Figure S1; Table S4). These results indicate that the observed phase-specific patterns did not reflect diel variation alone.
Additional time-course analyses identified 11 m/z features that increased between the first and second halves of the experiment in the neighbor condition. 6 of these were putatively assigned to hydrocarbons, organic acids and derivatives, organic nitrogen compounds, or organoheterocyclic compounds (Figures S2 and S4; Table S2). In the individual condition, only 1 m/z feature, putatively assigned to an organoheterocyclic compound, showed a comparable increase. These longer-term trends were distinct from the transient pre-peak increases, suggesting that volatile dynamics encompassed both progressive changes during plant development and phase-specific fluctuations associated with circumnutation.
The chemical classes tentatively annotated here, including benzenoids and organoheterocyclic compounds, are consistent with previously reported constitutive volatile emissions from P. sativum.10–15 Pea VOCs are also known to vary with developmental stage and plant organs.10–15 This inherent plasticity supports the biological plausibility of the temporal shifts we observed and indicates that volatile dynamics differed between individually grown and plants growing with a neighbor.
Collectively, these findings show that the presence of a neighbor was associated with changes in plant growth, movement dynamics, and volatile emission. Importantly, although the coordinated changes in circumnutation dynamics and volatile emissions observed during plant-plant interactions suggest a potential link between these processes, the present experiments do not allow us to establish a causal or mechanistic role for VOCs in modulating plant movement. Future experiments in which interacting plants are physically separated while one plant is exposed to VOC-containing airstreams, together with controls for other interaction cues such as root-derived emissions and changes in the light environment, will be necessary to disentangle the specific contribution of volatile emissions to circumnutation dynamics. Changes in VOC dynamics alone do not demonstrate communication between neighboring individuals. We therefore tested whether the circumnutation trajectories of neighboring plants exhibited enhanced temporal coupling using wavelet coherence analysis.
Neighboring plants exhibit greater movement coordination
To determine whether movement dynamics were coordinated between interacting plants, we compared wavelet coherence in real neighboring dyads with wavelet coherence in pseudo-dyads generated by randomly pairing plants from different dyads. Mean wavelet coherence was significantly higher in real dyads (mean = 0.381) than in pseudo-dyads (mean = 0.360 ± 0.0064). The observed value lay in the extreme right tail of the permutation-based null distribution (permutation test, p = 0.0004; Z = 3.31; Figure 3).
Figure 3. Neighboring plants exhibit stronger circumnutation synchronization than expected by chance.

(A) Representative wavelet coherence (WTC) map for a real plant dyad.
(B) WTC map for a pseudo-dyad generated by randomly pairing plants from different experimental dyads. Color indicates the magnitude of coherence between the two plant-movement time series as a function of normalized experimental time and frequency (mHz); warmer colors indicate higher coherence.
(C) Permutation test comparing the observed mean coherence of real dyads with a null distribution generated by randomly pairing plants across dyads (10,000 permutations). The histogram shows the distribution of group mean coherence values under the null hypothesis of no dyad-specific coupling. The dashed line indicates the null mean, dotted lines indicate ±1 standard deviation, and the solid line indicates the observed mean coherence of real dyads. The observed value falls outside the null distribution (p = 0.0004), indicating significantly greater coherence in real dyads than expected by chance.
These results indicate that neighboring plants exhibited greater time-frequency coordination in their movement than expected from random pairing. Because the pseudo-dyad analysis estimates coherence in randomly paired individuals from the same experimental condition, the increased coherence observed in real dyads reflects neighbor-dependent coupling rather than common external drivers. Nevertheless, wavelet coherence establishes temporal coordination, not the direction of influence between plants or the mechanism responsible for it.
The observed coordination is consistent with the concept of weakly coupled dynamical systems, in which interacting oscillators retain individual dynamics while exhibiting enhanced time-frequency alignment. Such systems can display entrainment or synchronization without behaving as a single integrated unit, as described by Kuramoto-type models.16,17 Several mechanisms may contribute to this coordination. The shared pot may have provided a pathway for belowground interactions capable of promoting coupling between plants.18,19 Mutual shade-avoidance responses may also contribute, because the movement of one plant could transiently alter the light environment perceived by its neighbor, including the red-to-far-red ratio.20 Airflow, humidity gradients, and other local microclimatic changes are additional non-mutually exclusive cues. These mechanisms were not isolated in the present study and should be addressed in future experiments.
These alternative mechanisms do not exclude a possible contribution of airborne emissions. The observed temporal relationship between volatile emissions and circumnutation is consistent with the possibility that interplant volatile emission contributes to exploratory growth during competition. However, the present design does not establish a causal link. The coordinated patterns observed here may arise from the combined influence of volatile, light-mediated, belowground, airflow, humidity, and other microclimatic cues rather than from a single release mechanism.
Neighbor presence is associated with shoot growth and node-specific leaf physiological responses
Morphological and physiological measurements provided additional evidence that neighbor presence was associated with a broader growth response. Plants grown with a neighbor exhibited higher daily growth rates, consistent with the enhanced shoot elongation observed in the morphological analysis (Figure S3; Tables S5 and S6). Increased shoot elongation is a common response to aboveground competition and may improve access to light.18–20 In climbing plants, greater shoot growth could also alter the spatial range of exploratory movement. Such elongation may additionally facilitate wider circumnutative excursions, increasing the probability of encountering a support.
Leaf physiological measurements obtained using a Dualex leaf-clip sensor revealed node-specific physiological differences between growth conditions (Table S5). Chlorophyll content was significantly higher in plants grown with a neighbor at the upper nodes n3 and n4 (Wilcoxon tests, p < 0.001 and p = 0.024, respectively), whereas differences at n1 and n2 were not statis-tically significant. Flavonol levels were higher in individually grown plants at basal nodes n1–n3 (p ≤ 0.010) but did not differ significantly at n4. Anthocyanin levels did not differ significantly between conditions at any node. The nitrogen balance index (NBI), in contrast, was consistently higher in plants grown with a neighbor across all four nodes (p ≤ 0.002).
Together, these patterns indicate spatially structured physio-logical plasticity in response to neighbor presence. Increased chlorophyll content in upper leaves of interacting plants may reflect enhanced photosynthetic investment in younger tissues positioned in more favorable light conditions under neighbor competition.20
Chemical dynamics and movement coordination define a framework for investigating plant-plant interactions
Neighbor presence was associated with changes in volatile emission, movement dynamics, and resource allocation in climbing pea plants. Plants grown with a neighbor showed faster growth, amplified circumnutations, and greater temporal coherence with their paired plant. These changes were accompanied by increased shoot allocation and greater chlorophyll content in the upper leaves, consistent with a competitive growth response.19 A set of volatile features also showed transient increases during the phase preceding the peak in circumnutation velocity, a pattern that was not detected in individually grown plants.
Further, the greater wavelet coherence observed in real dyads than in pseudo-dyads suggests that neighboring plants may behave as weakly coupled dynamical systems. Comparable forms of dynamical coordination21 occur in other biological systems, including locomotor interactions,22 circadian rhythms,23 neural activity,24 and synchronized behavior in social insects.25 However, the two main findings of the present study should be distinguished: wavelet coherence provides evidence of temporal coordination between the movements of neighboring plants, whereas the VOC analysis identifies an association between volatile and kinematic dynamics but does not establish that volatile cues caused this alignment.
A central limitation is that the experimental design did not separate airborne volatile compounds from belowground interactions, changes in light quality, airflow, humidity, or other microclimatic effects. Moreover, detecting volatile m/z features and observing their temporal association with movement cannot establish that the corresponding compounds are perceived by neighboring plants or function as evolved communicative signals. Future experiments should manipulate airborne exposure independently of physical and belowground contact, control airflow and microclimatic conditions, and test responses to chemically characterized compounds. Comparisons involving heterospecific neighbors could further determine whether these responses depend on neighbor identity.
In conclusion, neighbor presence was associated with coordinated changes in exploratory movement, volatile emission dynamics, shoot growth, and leaf physiology in P. sativum. The temporal association between volatile features and circumnutation, together with enhanced movement coherence in real neighboring dyads, provides a basis for investigating how volatile compounds relate to plant behavior. Establishing whether volatile compounds contribute causally to these responses will require experiments that distinguish their effects from alternative interaction pathways.
Resource Availability
Lead contact
Requests for further information and resources should be directed to, and will be fulfilled by, the lead contact, Bianca Bonato (bianca.bonato@unipd.it).
Materials availability
This study did not generate new, unique reagents.
Star★Methods
Key Resources Table
| REAGENT or RESOURCE | SOURCE | IDENTIFIER |
|---|---|---|
| Chemicals, peptides, and recombinant proteins | ||
| Benzene (≥99.9%), HPLC grade | Sigma-Aldrich, St.Louis, USA | 270709 |
| VOC calibration mixture. Composition: acetaldehyde, methanol, acetonitrile, acetone, isoprene, methyl ethyl ketone (2-butanone), benzene, toluene,x-xylene, a-pinene, 1,2,4-trimethylbenzene | Nippon Sanso Italia Sr.l. | (1 ppm in ultra-high-purity N2) |
| Murashige and Skoog Basal Salt Micronutrient Solution (10×, liquid) | Sigma Life Science, Milan, Italy | M0654-1L |
| Silica sand | Zanutta SPA, Padova, Italy | Type 16SS, 0.8–1.2 mm, S118 |
| Deposited data | ||
| Figures S1–S4 | supplemental information | Figures S1, S2, S3, and S4 |
| Tables S1, S2, S3, S4, S5, and S6 | supplemental information | Tables S1, S2, S3, S4, S5, and S6 |
| Experimental models: Organisms/strains | ||
| Snow pea (Pisum sativum var. saccharatum) | Fratelli Ingegnoli S.p.A., Milan, Italy | cv. Carouby de Maussane |
| Software and algorithms | ||
| Tofware | Tofwerk AG | Version 3.2.5 |
| Igor Pro (64-bit) | WaveMetrics | Version 9.05 |
| Python | Python Software Foundation | Version 3.12.7 |
| R statistical computing environment | R Foundation for Statistical Computing | Version 1.9.3 |
| MATLAB Wavelet Toolbox | MathWorks | Version R2024b |
| WinRHIZO 64 Pro | Regent Instruments Inc., Quebec, Canada | Version 2024a |
| Exact Mass Calculator | BioChemCalc | https://biochemcalc.com/exmc |
| OctoChemDB | EPFL | https://mstools.epfl.ch/monoisotopic/ |
| PubChem | National Center for Biotechnology Information (NCBI) | https://pubchem.ncbi.nlm.nih.gov/ |
| ClassyFire | Web-based application, Fiehn Lab, Wishart Research Group | https://cfb.fiehnlab.ucdavis.edu/ |
| Lotus | LOTUS initiative by Rutz et al.26 | https://lotus.naturalproducts.net/ |
| KEGG | Kanehisa Laboratories, Institute for Chemical Research (ICR), Kyoto University |
https://www.kegg.jp/ |
| ChEBI | EMBL-EBI, European Molecular Biology Laboratory, UK. | https://www.ebi.ac.uk/chebi/ |
| Custom image acquisition software | Ab.Acus S.r.l., Milan, Italy | ROOMORS-CAMRECORDER |
| Other | ||
| Vocus 2R time-of-flight mass spectrometer with AIM reactor | Tofwerk AG, Thun, Switzerland | P4Q-337 |
| Dualex leaf-clip sensor | METOS, Pessl Instruments | N/A |
| EPSON scanner | Seiko Epson Corporation, Suwa, Japan | Epson LA2400 Expression 13000XL |
| Precision caliper, 0.01 mm resolution | Vetrotecnica S.r.l | N/A |
| Precision balance | Radwag, Radom, Poland | AS X2 PLUS |
| Filter paper | Cytiva | Whatman 3MM Chr |
| Growth chamber phytotron | Esse Costruzioni S.r.l., Frosinone, Italy | Custom-built |
| RGB-infrared camera | Reolink | RLC-833A |
| LED growth lamp | Ledvance, Germany | HB P 200W 840 110DEG IP66 PS |
| Photometer and probe | Delta Ohm, Italy | HD2302.0; LP471PAR |
| PoE network switch | Ubiquiti Networks | UniFi Pro, 24-port |
| Vacuum pump | Ebara | PDV 500 |
| Perfluoroalkoxy (PFA) tubing | Nordival Srl-Swagelok Italia, Brescia, Italy | 3/8-inch outer diameter. PFA-350 |
| Defender flow meter | Mesa Labs | Defender 500 M |
Experimental Model and Study Participant Details
Snow pea seeds (Pisum sativum var. saccharatum cv. Carouby de Maussane) were obtained from Fratelli Ingegnoli S.p.A. (Milan, Italy). Plants were grown under controlled environmental conditions in a custom-built phytotron, as described in the STAR Methods section. The study included individual plants and plants grown in neighboring pairs. The individual-plant condition included 14 biological replicates, whereas the neighbor condition included 14 dyads (28 plants). The plants’ developmental stage at the time of the experiments was 4 days of germination and 14 days of data collection after the transplant.
No human participants, animals, cell lines, primary cell cultures, or microbial strains were used in this study. Institutional approval for plant experimentation was not applicable.
Method Details
Biological materials and growth conditions
Snow pea seeds (Pisum sativum var. saccharatum cv. Carouby de Maussane; Fratelli Ingegnoli S.p.A., Milan, Italy) were imbibed in tap water for 90 min and germinated in darkness at 22°C for 4 days on strips of filter paper (Whatman 3MM Chr, Cytiva, United States) saturated with demineralized water. Seeds were positioned 1.5 cm apart and 0.5 cm from the top of each strip, with the hilum and micropyle oriented downward. Uniform seedlings were selected according to root and shoot length and transferred to a custom-built phytotron (Esse Costruzioni S.r.l., Frosinone, Italy) for subsequent growth.
The phytotron consisted of a climate-controlled room in which temperature, relative humidity, light intensity, irrigation, fertilization, and air recirculation were regulated through a central control panel. Air quality and environmental stability were maintained using a recirculating air-treatment system comprising a prefilter, HEPA filter, fan, and steam humidifier. A slight positive pressure (5 Pa relative to the surrounding environment) was maintained in each growth chamber. Growth chambers had a volume of 1 m3 and contained cylindrical pots filled with silica sand (Type 16SS, particle size 0.8–1.2 mm; density 1.4 g/cm3).
The experiment included an individual condition and a neighbor condition. 14 biological replicates were used in the individual condition, with one seedling per pot in pots measuring 20 cm in diameter and 20 cm in height. The neighbor condition included 14 dyads (28 plants), with two seedlings planted in each 40-cm-diameter, 20-cm-high pot on opposite sides of a central support (1.2 cm in diameter and 54 cm in height). Each seedling was positioned 10 cm from the support. The substrate volume in neighbor-condition pots was approximately twice that used in individual-condition pots, providing a comparable substrate volume per plant. Representative recordings of the individual and neighbor conditions are provided in Videos S1 and S2, respectively.
Seedlings were grown for 14 days after transplantation. Pots were irrigated and fertilized every other day using half-strength Murashige and Skoog Basal Salt Micronutrient Solution (10×, liquid, plant cell culture tested; Sigma Life Science, Milan, Italy) prepared with microfiltered water. Water and nutrient-solution volumes were adjusted according to the number of plants to provide a comparable amount per plant across conditions. Growth chambers were maintained at 22 ± 1.5°C and 40 ± 10% relative humidity under an 11.25-h photoperiod, with lights on from 05:45 to 17:00. Cool-white LED lamps (HB P 200W 840 110DEG IP66 PS, Ledvance, Germany) were positioned 50 cm above the seedlings. Photosynthetic photon flux density was 400 μmol m-2 s-1 at seedling height, measured using an HD2302.0 photometer and LP471PAR probe (Delta Ohm, Italy).
One chamber contained an unplanted pot filled with the same substrate and served as a blank control for background environmental signals. Measurements from this chamber were subtracted from experimental measurements during VOC data processing.
Each growth chamber was equipped with two adjustable-rail-mounted 4K UHD RGB-infrared cameras (RLC-833A, Reolink). Cameras were powered through Ethernet and connected to a 24-port UniFi Pro switch (Ubiquiti Networks). Image acquisition and real-time monitoring were performed on a dedicated workstation using custom software (ROOMORS-CAMRECORDER, Ab.Acus S.r.l., Milan, Italy). Image acquisition and mass-spectrometry timestamps were synchronized automatically using the Network Time Protocol (NTP), with both the mass spectrometer and workstation connected to the internet.
Morphological sampling and analysis
Seedlings were harvested 14 days after transplantation into the growth chambers. Aboveground and belowground morphological measurements were performed using a precision caliper with 0.01-mm resolution. The entire root system and stem were then scanned using an EPSON scanner (LA2400 scanner, Epson Expression 13000XL, Seiko Epson Corporation, Suwa, Japan), and root and stem morphological traits were quantified using WinRHIZO 64 Pro 2024a (Regent Instruments Inc., Québec, QC, Canada). Fresh biomass of roots, shoots, and seeds was measured using a precision balance (RADWAG, Radom, Poland). Plant tissues were subsequently dried at 70°C for at least 72 h to determine dry biomass.
Aboveground traits included shoot length, tendril- and branch-node lengths, internode lengths from i0 to i4, epicotyl length, and fresh and dry shoot biomass. Belowground traits included primary root length, mean root diameter, and fresh and dry root biomass (Figure S3). The complete list of morphological variables is provided in the Tables S5 and S6. Each plant was assigned to the individual or neighbor condition.
Optical assessment of leaf physiological parameters
Leaf physiological parameters were measured using a Dualex leaf-clip sensor (METOS, Pessl Instruments), a non-destructive optical instrument for assessing leaf biochemical indices. The Dualex uses dual excitation and detection wavelengths to estimate leaf pigment levels from chlorophyll fluorescence and transmittance. Measurements were conducted at the end of the experimental period while plants were alive in their sand-filled pots in the respective growth chambers. The sensor was calibrated before each use according to the manufacturer’s instructions.
Measurements were taken from fully expanded, light-exposed leaves without removing tissue. Leaves were measured sequentially from the oldest (lowest) to the youngest (uppermost), capturing variation across leaf developmental stages. The recorded indices were the Chlorophyll Index, Flavonol Index, Anthocyanin Index, and Nitrogen Balance Index (NBI), calculated as the ratio of chlorophyll to flavonols. These indices were used as optical indicators of leaf pigment status and physiological condition. The complete list of recorded variables is provided in Tables S5 and S6. Each plant was assigned to the individual or neighbor condition.
Three-dimensional kinematic data acquisition and analysis
Stereo images of seedlings were acquired continuously from all eight growth boxes for 14 days, at a rate of one image every 3 min (0.00556 Hz). No corrupted or missing images were reported during acquisition.
Stereo image-derived time series were analyzed to compare plant movement between the individual and neighbor conditions. The apex parameters examined were circumnutation velocity, maximum acceleration, minimum acceleration, circumnutation path length, and area explored during circumnutation. The first 3 days of data were excluded from the kinematic analysis so that analysis began around the emergence of the first leaves.
Time series of apex circumnutation parameters obtained through the tracking process27,28 were processed in Python 3.12.7. Daily growth rates were calculated by differencing cumulative apex displacement from the origin at 24-h intervals. Each circumnutation-event timestamp was converted to minutes from the start of the experiment to establish a common temporal reference across datasets. Preprocessed data, including daily growth rates and apex circumnutation features, were organized in long format using consistent plant, condition, and time identifiers and exported as comma-separated files for statistical analysis.
Wavelet analysis of real and pseudo-dyads
Wavelet coherence between the horizontal (x-coordinate) trajectories of the two plants in each dyad was calculated in MATLAB using the Wavelet Toolbox, following an approach similar to Bonato et al.8 A continuous wavelet transform was used to estimate coherence over periods ranging from 12 min to 6 h. Regions affected by edge effects were excluded using the cone of influence.
For each real dyad, mean coherence was calculated by averaging coherence values across time and the selected period band after masking regions outside the cone of influence.
Volatile organic compound sampling and analysis
Volatile organic compounds emitted by seedlings were monitored using a high-resolution time-of-flight mass spectrometer (Vocus 2R TOF-MS; Tofwerk AG, Thun, Switzerland) installed inside the phytotron. The instrument supported either a proton-transfer-reaction (PTR) or an adduct-ionization-mechanism (AIM) reactor. The AIM reactor was selected because comparative analyses indicate that this ionization approach detects a greater number of distinct VOC signals than PTR.29
Benzene (C6H6+) was used as the primary photoabsorbent reagent-ion source and delivered at 0.25 L min-1. Gaseous benzene (C6H6, Sigma Aldrich, St. Louis, MO, USA, ≥99.9%) was maintained in a heated permeation tube, introduced into a vacuum-ultraviolet lamp housing, and photoionized to generate C6H6+ ions and photoelectrons.30 The permeation-tube temperature was set to 80°C, the AIM-reactor temperature to 50°C, and the ionization-source pressure to 50 mbar.
Instrument sensitivity and mass calibration were performed before data acquisition using a reference VOC mixture (1 ppm in ultra-high-purity N2) containing acetaldehyde, methanol, acetonitrile, acetone, isoprene, methyl ethyl ketone (2-butanone), benzene, toluene, x-xylene, a-pinene, 1,2,4-trimethylbenzene. The calibration mixture was delivered at 0.01 L min-1. A multiport valve (Tofwerk AG) connected the reactor inlet to multiple sampling lines. Each growth chamber was connected to a dedicated valve port using 6 m of perfluoroalkoxy tubing with a 3/8-inch outer diameter. Samples were continuously drawn at 2 L min-1 using a Defender 500 M flow meter (Mesa Labs). The valve switched between chambers every 7 min, providing at least one acquisition per hour for each chamber.
At the reactor exit, product ions passed through a 1-mm orifice into a radio-frequency-only quadrupole ion guide. Neutral molecules were removed using an Ebara PDV 500 vacuum pump. Ions were transferred under controlled pressures of 10-2 mbar in the quadrupole and 10-5 mbar in the lens stack before orthogonal extraction into the TOF analyzer, which operated below 10-6 mbar. Mass spectra were acquired over m/z 0–250 at a mass resolving power of 10,000–11,000.31 Data were recorded at 5-s temporal resolution, with timestamps stored in Coordinated Universal Time.
Vocus data preprocessing
Raw AIM-TOF-MS data were preprocessed using Tofware 3.2.5 (Tofwerk), running in Igor Pro 9.05 (64-bit). Processing included spectrum definition, refinement of peak width and shape, mass calibration, peak detection, and data export. The exported comma-separated files contained one row per time point and one column per detected m/z feature. Files were merged using a custom Python module developed for this study (Python 3.12.7).
For each acquisition cycle, the first 2 min and final 1 min were removed to minimize transition effects between sampled chambers, and mean feature intensity was calculated over the remaining interval. After acquisition cycles were merged, the resulting time series contained one measurement approximately every 56 min for each detected m/z feature in each chamber. Each feature was normalized to the corresponding benzene reagent-ion intensity. Background contributions were removed by subtracting the intensity measured in the blank chamber at each time point. Background-corrected time series were restricted to the period following the experiment-specific start time to align temporal records across experiments. Only features within the m/z range 30–250 were retained. Features with pairwise correlations greater than 99% were removed; these highly correlated features predominantly represented isotopes of monoisotopic signals.
Annotation of m/z features
Candidate elemental formulas for selected m/z features were queried using OctoChemDB (https://mstools.epfl.ch/monoisotopic/).32 Candidate formulas were generated using the measured masses and ionization information, with a mass-accuracy threshold of 15 ppm. Theoretical exact masses were calculated using Exact Mass Calculator (https://biochemcalc.com/exmc), which accounts for the charge of the elemental formula. Exact-mass error was calculated by comparing measured and theoretical masses as follow:
Candidate compounds consistent with the elemental formulas were identified through a manually curated search of PubChem (https://pubchem.ncbi.nlm.nih.gov/).33 InChIKeys34 were retrieved where available and used to obtain classifications from ClassyFire (https://cfb.fiehnlab.ucdavis.edu/). Candidate compounds were manually assigned to putative chemical classes: benzenoids, hydrocarbons, organic acids and derivatives, organic nitrogen compounds, organic oxygen compounds, organoheterocyclic compounds, or unknown. Where available, Lotus database (https://lotus.naturalproducts.net/) information, KEGG entry codes (https://www.kegg.jp/), ChEBI (https://www.ebi.ac.uk/chebi/) links, and putative parent compounds and their classifications were also recorded.
Because direct-injection AIM-TOF-MS does not provide chromatographic separation and cannot distinguish structural isomers, the resulting annotations represent putative assignments rather than definitive compound identifications. See Figure S4.
Quantification and Statistical Analysis
Statistical software and general approach
Statistical analyses were conducted in R using the mgcv package (version 1.9.3), unless otherwise specified. Data preprocessing and organization were performed in Python 3.12.7. Wavelet coherence and permutation analyses were performed in MATLAB using the Wavelet Toolbox. Statistical tests were two-sided unless a directional hypothesis was specified. Multiple comparisons among m/ z features were controlled using the Benjamini–Hochberg false discovery rate (FDR) procedure, with a significance threshold of α = 0.05.
Biological replication was defined by experimental condition: 14 individual plants in the individual condition and 14 dyads (28 plants) in the neighbor condition. For analyses in which plants within a dyad were averaged, the dyad/pot was treated as the experimental unit. Repeated time points from the same plant or chamber were not treated as independent biological replicates.
Morphological and leaf physiological comparisons
Morphological traits and Dualex-derived physiological indices were compared between individual and neighbor conditions using two-sided Wilcoxon rank-sum tests. For each trait, the condition associated with higher trait expression was identified by comparing the group medians. The complete lists of measured morphological and physiological variables are provided in Tables S5 and S6 and Figure S3.
Temporal dynamics of kinematic parameters
Temporal changes in apex kinematics were analyzed using generalized additive mixed models (GAMMs) implemented in the R package mgcv (version 1.9.3). Separate models were fitted for circumnutation velocity, maximum acceleration, minimum acceleration, circumnutation path length, area explored during circumnutation, and daily growth rate, as applicable.
Each response was modeled as a function of experimental condition and time from experiment onset. Nonlinear temporal trajectories were represented using smooth functions of time, with condition-specific smooths used to estimate potentially different temporal dynamics between conditions. Plant identity was incorporated through plant-specific smooth deviations to account for repeated measurements. Within-plant temporal autocorrelation was modeled using a first-order autoregressive (AR1) structure. Models were fitted using restricted maximum likelihood (REML).
Differences in temporal dynamics between conditions were tested by comparing models with a shared temporal smooth against models with condition-specific smooths. These comparisons were fitted using maximum likelihood (ML), and likelihood-ratio tests were used to assess the evidence for condition-dependent temporal trajectories.
Screening and temporal analysis of m/z features
One-sample t-tests were used to identify m/z features with mean intensity significantly greater than zero across experimental replicates. Temporal increases were assessed using paired, one-sided t-tests comparing the second half of the experiment with the first half and comparing the pre-peak phase with both the baseline and peak phases (see Figure S2). Multiple testing across features was controlled using the Benjamini–Hochberg FDR procedure, with α = 0.05.
Wavelet coherence and permutation testing
Wavelet coherence between the horizontal trajectories of plants within real dyads was summarized by averaging coherence across time and periods from 12 min to 6 h, after excluding regions outside the cone of influence. A null distribution was generated by randomly re-pairing plants into pseudo-dyads and calculating group-level mean coherence for each of 10,000 permutations.
The empirical significance of the observed group mean coherence was evaluated by comparing it with the permutation distribution. The permutation-based p value was calculated as the proportion of permutations whose group mean coherence was greater than or equal to the observed mean for the real dyads.
Relating VOC dynamics to kinematics
Feature-level summaries were calculated over the full experimental period and within temporal windows defined using the kinematic results. Circumnutation apex velocity increased approximately 7 days after experiment onset (approximately 10,080 min; Figure 1). Three phases were therefore defined: baseline, from experiment onset to day 4; pre-peak, from day 4 to day 6; and peak, from day 6 to day 8 (Figure 2). For each m/z feature, mean intensity was calculated over the full-time series, separately for the first and second halves of the experiment (Figure S2), and within each defined phase. These summaries were used for subsequent statistical comparisons. For the joint representation of VOC and movement, the VOC time series and the apex circumnutation time series of each replicate were placed on a common grid of 60-minute bins expressed as minutes from the start of the experiment, using the shared time reference established by network time synchronization between the mass spectrometer and the imaging system. Within each chamber, apex velocity was averaged across the two plants of the pot so that both measurements refer to the same experimental unit. Group means across the 14 replicates were smoothed with a centred 6-hour moving average for display only; no smoothing was applied to any data entering a statistical test.
Photoperiod control analysis
To assess whether the VOC dynamics were confounded by diel regulation, each acquisition time point was classified as belonging to the light or the dark period according to the set photoperiod (lamps on 05:45-17:00, 11.25 h light per 24 h cycle). Note that the three phase windows are integer multiples of 24 h and therefore each span complete light-dark cycles. Day/night differences were tested for each of the seven signals by comparing every replicate’s mean light-period signal with its mean dark-period signal using paired two-sided t-tests, corrected using FDR (α = 0.05). To control for photoperiod explicitly, phase means were then recomputed for each replicate as a weighted average of its light-period and dark-period means, with the weights fixed at the photoperiod fraction (0.469 light, 0.531 dark) for all three phases; because the weights are identical across phases, differences in the light/dark composition of a phase cannot contribute to a contrast. The pre-peak versus baseline and pre-peak versus peak contrasts were repeated on these standardized means, with the same paired one-sided tests and FDR correction used for the main analysis (see Figure S1).
Supplementary Material
Supplemental information can be found online at https://doi.org/10.1016/j.cub.2026.10.050.
Highlights.
Volatile features rise before peak circumnutation velocity in pea plants
Neighboring pea plants show more synchronized circumnutation than random pairs
Neighbor presence enhances circumnutation speed and spatial exploration
Neighbor presence promotes shoot growth and alters leaf physiology
In brief.
Bonato et al. combine three-dimensional kinematic tracking, real-time volatilomic profiling, wavelet coherence, and morphological analyses to investigate plant-plant interactions. In neighboring pea plants, volatile features increase before peak circumnutation velocity, while paired plants show enhanced temporal coordination, revealing a temporal association between chemical dynamics and exploratory movement.
Acknowledgments
This research was funded by the European Union (ERC, ROOMors, grant number 101096728 to U.C.). The views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.
Footnotes
Author Contributions
Conceptualization, U.C.; methodology, U.C., B.B., V.S., and S.A.; formal analysis, V.S.; investigation, B.B., S.A., G.G., and S.G.; data curation, V.S., B.B., S.A., G.G., and S.G.; writing – original draft, B.B.; writing – review and editing, U.C., B.B., V.S., S.A., M.D., L.R., and S.G.; supervision, U.C.; funding acquisition, U.C.; and data interpretation, B.B., V.S., S.A., and U.C. All authors approved the final version of the manuscript.
Declaration of Interests
The authors declare no competing interests.
Data and code availability
Data for this manuscript have been deposited at Zenodo repository at https://doi.org/10.5281/zenodo.23388978 and are publicly available as of the date of publication.
All original code has been deposited at Zenodo repository and is publicly available at https://doi.org/10.5281/zenodo.23388978 as of the date of publication.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.
References
- 1.Stolarz M. Circumnutation as a visible plant action and reaction: physiological, cellular and molecular basis for circumnutations. Plant Signal Behav. 2009;4:380–387. doi: 10.4161/psb.4.5.8293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Guerra S, Peressotti A, Peressotti F, Bulgheroni M, Baccinelli W, D’Amico E, Gómez A, Massaccesi S, Ceccarini F, Castiello U. Flexible control of movement in plants. Sci Rep. 2019;9:16570. doi: 10.1038/s41598-019-53118-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mugnai S, Azzarello E, Masi E, Pandolfi C, Mancuso S. In: Rhythms in Plants: Dynamic Responses in a Dynamic Environment. Mancuso S, Shabala S, editors. Springer International Publishing; 2015. Nutation in plants; pp. 19–34. [DOI] [Google Scholar]
- 4.Wooten MB, Walker ID. Vine-inspired continuum tendril robots and circumnutations. Robotics. 2018;7:58. doi: 10.3390/robotics7030058. [DOI] [Google Scholar]
- 5.Guerra S, Bonato B, Wang Q, Peressotti A, Peressotti F, Baccinelli W, Bulgheroni M, Castiello U. Kinematic evidence of root-to-shoot signaling for the coding of support thickness in pea plants. Biology. 2022;11:405. doi: 10.3390/biology11030405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Bonato B, Simonetti V, Bulgheroni M, Wang Q, Guerra S, Quaggiotti S, Ruperti B, Castiello U. Evidence of motor intentions in plants: A kinematical study. J Comp Psychol. 2023;137:228–237. doi: 10.1037/com0000351. [DOI] [PubMed] [Google Scholar]
- 7.Bonato B, Wang Q, Guerra S, Simonetti V, Bulgheroni M, Quaggiotti S, Ruperti B, Castiello U. ‘United we stand, divided we fall’: intertwining as evidence of joint actions in pea plants. AoB Plants. 2024;16:plad088. doi: 10.1093/aobpla/plad088. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bonato B, Simonetti V, Castiello U. Revealing synchrony in pea plants using wavelet coherence analysis. Sci Rep. 2025;15:36226. doi: 10.1038/s41598-025-20198-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Simonetti V, Ravazzolo L, Ruperti B, Quaggiotti S, Castiello U. The effects of a neighbour and its identity on roots’ plastic growth. AoB Plants. 2025;17:plaf031. doi: 10.1093/aobpla/plaf031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dudareva N, Negre F, Nagegowda DA, Orlova I. Plant volatiles: recent advances and future perspectives. Crit Rev Plant Sci. 2006;25:417–440. doi: 10.1080/07352680600899973. [DOI] [Google Scholar]
- 11.Baldwin IT. Plant volatiles. Curr Biol. 2010;20:R392–R397. doi: 10.1016/j.cub.2010.02.052. [DOI] [PubMed] [Google Scholar]
- 12.Ninkovic V, Markovic D, Rensing M. Plant volatiles as cues and signals in plant communication. Plant Cell Environ. 2021;44:1030–1043. doi: 10.1111/pce.13910. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Avesani S, Castiello U, Ravazzolo L, Bonato B. Volatile organic compounds in pea plants: a comprehensive review. Front Plant Sci. 2025;16:1591829. doi: 10.3389/fpls.2025.1591829. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ceballos R, Fernández N, Zúñiga S, Zapata N. Electrophysiological and behavioral responses of pea weevil Bruchus pisorum L. (Coleoptera: Bruchidae) to volatiles collected from its host Pisum sativum L) Chilean J Agric Res. 2015;75:202–209. doi: 10.4067/S0718-58392015000200009. [DOI] [Google Scholar]
- 15.Thöming G, Norli HR, Saucke H, Knudsen GK. Pea plant volatiles guide host location behaviour in the pea moth. Arthropod-Plant Interact. 2014;8:109–122. doi: 10.1007/s11829-014-9292-5. [DOI] [Google Scholar]
- 16.Pikovsky A, Rosenblum M, Kurths J. Synchronization: A Universal Concept in Nonlinear Sciences. Cambridge University Press; 2010. [DOI] [Google Scholar]
- 17.Acebrón JA, Bonilla LL, Pérez Vicente CJ, Ritort F, Spigler R. The Kuramoto model: A simple paradigm for synchronization phenomena. Rev Mod Phys. 2005;77:137–185. doi: 10.1103/RevModPhys.77.137. [DOI] [Google Scholar]
- 18.Zahra N, Wahid A, Hafeez MB, Waqas M, Kausar A, Alsahli AA. Plant growth promoters (PGPs) alleviate salt stress, enhance physiological parameters, secondary metabolites and silymarin content in milk thistle. J Soil Sci Plant Nutr. 2025;25:948–970. doi: 10.1007/s42729-024-02177-y. [DOI] [Google Scholar]
- 19.Schell V, Kervroëdan L, Depeaux G, Delzon S, Faucon MP. Unravelling the phenotypic plasticity of functional traits in relay-cropping: the case of soybean with winter barley. Plant Soil. 2025;513:1141–1158. doi: 10.1007/s11104-025-07242-8. [DOI] [Google Scholar]
- 20.Casal JJ. Shade avoidance. Arabidopsis Book. 2012;10:e0157. doi: 10.1199/tab.0157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Strogatz S. Synchronization: A universal concept in nonlinear sciences. Phys Today. 2003;56:47. doi: 10.1063/1.1554136. [DOI] [Google Scholar]
- 22.Doering GN, Lee CL, Dalnoki-Veress K. Synchronized locomotion can improve spatial accessibility inside ant colonies. Proc R Soc B. 2023;290:20231805. doi: 10.1098/rspb.2023.1805. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ramkisoensing A, Meijer JH. Synchronization of biological clock neurons by light and peripheral feedback systems promotes circadian rhythms and health. Front Neurol. 2015;6:128. doi: 10.3389/fneur.2015.00128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Elson RC, Selverston AI, Huerta R, Rulkov NF, Rabinovich MI, Abarbanel HD. Synchronous behavior of two coupled biological neurons. Phys Rev Lett. 1998;81:5692–5695. doi: 10.1103/PhysRevLett.81.5692. [DOI] [Google Scholar]
- 25.Klibaite U, Shaevitz JW. Paired fruit flies synchronize behavior: uncovering social interactions in Drosophila melanogaster. PLoS Comput Biol. 2020;16:e1008230. doi: 10.1371/journal.pcbi.1008230. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Rutz A, Sorokina M, Galgonek J, Mietchen D, Willighagen E, Gaudry A, Graham JG, Stephan R, Page R, Vondrásek J, et al. The LOTUS initiative for open knowledge management in natural products research. eLife. 2022;11:e70780. doi: 10.7554/eLife.70780. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Simonetti V, Bulgheroni M, Guerra S, Peressotti A, Peressotti F, Baccinelli W, Ceccarini F, Bonato B, Wang Q, Castiello U. Can plants move like animals? A three-dimensional stereovision analysis of movement in plants. Animals. 2021;11:1854. doi: 10.3390/ani11071854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Simonetti V, Bonato B, Guerra S, Avesani S, Semenzato L, Bulgheroni M, Gjinaj G, Ravazzolo L, Dadda M, Castiello U. Mind(the)Plant: An expandable multimodal facility for the integrated characterization of plant behaviour. Methods Ecol Evol. 2026:1–17. doi: 10.1111/2041-210x.70411. 00. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Avesani S, Bonato B, Simonetti V, Guerra S, Ravazzolo L, Gjinaj G, Dadda M, Castiello U. Comparing proton transfer reaction (PTR) and adduct ionization mechanism (AIM) for the study of volatile organic compounds. Molecules. 2026;31:402. doi: 10.3390/molecules31030402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Riva M, Pospisilova V, Frege C, Perrier S, Bansal P, Jorga S, Sturm P, Thornton JA, Rohner U, Lopez-Hilfiker F. Evaluation of a reduced-pressure chemical ion reactor utilizing adduct ionization for the detection of gaseous organic and inorganic species. Atmos Meas Tech. 2024;17:5887–5901. doi: 10.5194/amt-17-5887-2024. [DOI] [Google Scholar]
- 31.Lavi A, Vermeuel MP, Novak GA, Bertram TH. The sensitivity of benzene cluster cation chemical ionization mass spectrometry to select biogenic terpenes. Atmos Meas Tech. 2018;11:3251–3262. doi: 10.5194/amt-11-3251-2018. [DOI] [Google Scholar]
- 32.Silvestre R, Martinent R, Menin L, Gasilova N, Mutel V, Portmann C, Patiny L. OctoChemDB: An aggregated database for small molecule identification using high-resolution MS data. Anal Chem. 2026;98:6102–6108. doi: 10.1021/acs.analchem.5c06761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wang Y, Xiao J, Suzek TO, Zhang J, Wang J, Bryant SH. PubChem: A public information system for analyzing bioactivities of small molecules. Nucleic Acids Res. 2009;37:W623–W633. doi: 10.1093/nar/gkp456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Djoumbou Feunang Y, Eisner R, Knox C, Chepelev L, Hastings J, Owen G, Fahy E, Steinbeck C, Subramanian S, Bolton E, et al. ClassyFire: automated chemical classification with a comprehensive, computable taxonomy. J Cheminform. 2016;8:61. doi: 10.1186/s13321-016-0174-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data for this manuscript have been deposited at Zenodo repository at https://doi.org/10.5281/zenodo.23388978 and are publicly available as of the date of publication.
All original code has been deposited at Zenodo repository and is publicly available at https://doi.org/10.5281/zenodo.23388978 as of the date of publication.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.
