Motor cortex rTMS modulates adult ocular dominance plasticity Authors: İzel D Sarı, Vasileia Papakosta, Xavier Corominas Teruel, Hanna Willis, Antoni Valero-Cabré, Claudia Lunghi Published: 2026 ABSTRACT AbstractWell-established evidence supports cross-modal interactions of sensory brain plasticity. Here, we extend cross-modality beyond the boundaries of sensory regions and provide first-time evidence in adult healthy humans that motor plasticity influences homeostatic visual plasticity. To this end, we employed repetitive transcranial magnetic stimulation (rTMS), known to modulate the excitation/inhibition balance in the primary motor cortex (M1), while examining the impact of this intervention on ocular dominance plasticity induced by short-term monocular deprivation (MD). Specifically, participants completed, in counterbalanced order, two neuromodulation sessions delivering either Active or Sham rTMS (10 Hz, 1600 pulses), during a 90-min MD period. We show that Active but not Sham rTMS significantly increased amplitudes of peak-to-peak motor evoked potentials, confirming facilitatory cortico-spinal after-effects driven by M1 stimulation. While robust ocular dominance shifts were observed for both rTMS conditions, greater increases of cortico-spinal excitability scaled with smaller ocular dominance shifts, suggesting interference of primary visual cortex plasticity ensuing a modulation of M1 excitability and competitive interactions between motor and visual plasticity. These results hint at global homeostatic mechanisms governing adult cortical plasticity across distant networks and highlight the need to consider cross-modal interactions between motor and visual systems when designing learning and rehabilitation strategies for adults.Keywords: monocular deprivation, neuroplasticity, transcranial magnetic stimulation, visual plasticity Introduction Despite a marked decrease in sensory cortex neuroplasticity in adulthood (Berardi et al. 2000; Pascual-Leone et al. 2005), ocular dominance plasticity in the adult visual cortex following short-term monocular deprivation (MD) has been robustly demonstrated (Castaldi et al. 2020; Baroncelli and Lunghi 2021; Min 2025). Briefly depriving one eye of normal vision causes a transient ocular dominance shift favoring the deprived eye, reflecting residual homeostatic plasticity in adults. This change, initially measured behaviorally (Lunghi et al. 2011, 2013; Zhou et al. 2013; Bai et al. 2017; Prosper et al. 2023), is reflected by changes in primary visual cortex (V1) activity, whereby after 2–2.5 h of MD, responses relayed to V1 by the deprived eye increase whereas those conveyed by its non-deprived counterpart decrease (Zhou et al. 2015; Lunghi et al. 2015a; Chadnova et al. 2017; Binda et al. 2018). Importantly, this shift is accompanied by a decrease in resting gamma aminobutyric acid (GABA) concentration in the early visual cortex (Lunghi et al. 2015b), highlighting the importance of excitation/inhibition (E/I) balance for adult ocular dominance plasticity. It is well known that the motor cortex preserves a high plastic potential throughout life, as exemplified by motor learning paradigms (Karni et al. 1995), and the ability of adult brains to retain new motor skills over time (Boyke et al. 2008; Berghuis et al. 2016; Mooney et al. 2017, 2019). This ability relies on plasticity throughout tightly interconnected motor and sensorimotor cortices (Dayan and Cohen 2011), mediated by shifts in inhibition (Bachtiar and Stagg 2014). Studies in animal models have shown that long-term potentiation-like processes underlie motor skill acquisition (Rioult-Pedotti et al. 1998, 2000; Harms et al. 2008), while pointing to reduced GABAergic signaling as an essential mechanism for motor learning (Jacobs and Donoghue 1991; Castro-Alamancos et al. 1995; Trepel and Racine 2000). In humans, pharmacological interventions (Bütefisch et al. 2000; Ziemann et al. 2001), noninvasive brain stimulation (Liepert et al. 1998; Perez et al. 2004; Rosenkranz et al. 2007; Coxon et al. 2014), and magnetic resonance spectroscopy (Sampaio-Baptista et al. 2015; Li et al. 2024) support the role of reduced GABAergic inhibition in the primary motor cortex (M1) facilitating motor learning. To this regard, studies have shown that baseline GABA levels predict motor learning performance (Stagg et al. 2011; Coxon et al. 2014; Kim et al. 2014; King et al. 2020; Li et al. 2024), whereas reductions in GABA track learning-related improvements (Floyer-Lea et al. 2006; Sampaio-Baptista et al. 2015; Kolasinski et al. 2019; Bell et al. 2023; Hehl et al. 2025). Although both visual and motor plasticity phenomena may rely on similar general changes of intracortical E/I balance, potential interactions and influences between these processes remain poorly understood. Healthy adult human studies exploring the two concomitantly have mostly employed visuomotor paradigms, embedding these two components within the same task (Ostry and Gribble 2016). By showing that visual plasticity induced by short-term MD is reduced during motor sequence learning (Sarı and Lunghi 2023), our team has recently provided proof of interactions between visual and motor plasticity phenomena in adult humans, putting forward the notion that motor plasticity may interfere with homeostatic visual plasticity. Moreover, this finding led us to hypothesize the existence of a global homeostatic regulation mechanism involving multiple brain regions, engaged to avoid instability when intense local shifts of E/I balance occur. Here, we tested whether directly modulating the excitability of the M1 region by means of repetitive transcranial magnetic stimulation (rTMS) could affect the propensity and magnitude of well-established visual homeostatic plasticity phenomena induced by short-term MD in healthy adult humans. To determine whether disruption of the motor cortex E/I balance influences visual plasticity, we delivered either Active or Sham high-frequency rTMS (10 Hz, 1600 pulses, for 28 min) to the motor cortex in two separate, order-counterbalanced sessions. This stimulation paradigm is a well-established conventional protocol (Maeda et al. 2000a, 2002; Gangitano et al. 2002; Fierro et al. 2005; Rounis et al. 2005; Fitzgerald et al. 2006; Valero-Cabré et al. 2017) known to transiently enhance cortical excitability in ~70%–90% of the tested participants. We then measured ensuing changes of visual plasticity by monitoring shifts of ocular dominance following a 90-min period of MD intervention. Based on previous findings suggesting competition between motor and visual plasticity phenomena (Sarı and Lunghi 2023), we hypothesized: (i) higher plastic changes of ocular dominance induced by MD after Active compared to Sham rTMS; whereas considering interindividual variability in the net modulatory outcomes induced by high frequency rTMS, we also predicted (ii) higher modulations in M1 excitability scaling with lower ocular dominance plasticity measurements. Materials & methods Participants An initial cohort of n = 28 neurologically and psychologically healthy volunteers (12F, 16 M, mean age = 24.8, SD = 4.8) with normal or corrected-to-normal vision were recruited for the study. Baseline data were screened for task stability and physiological comparability before analysis. For binocular rivalry, ocular dominance estimates from the baseline blocks were inspected to confirm that participants produced a stable dominance estimate across the task. Participants were excluded if the dominance index differed by more than 3-fold across the four sub-blocks within either pre-stimulation baseline condition (associated with Sham or Active rTMS sessions), resulting in the exclusion of one participant. MEP recruitment curves were then screened to ensure that Sham and Active baseline input–output (I/O) functions were also comparable in profile. Participants were excluded when the two curves showed a qualitative recruitment-profile mismatch that could not be accounted for by plausible resting motor threshold (RMT) misestimation or by a simple vertical difference in motor evoked potential (MEP) amplitude response gain. Such mismatches included crossing of the fitted curves within the tested RMT range, saturation in one of the rTMS condition baselines while the other continued to rise approximately linearly, or largely non-overlapping response ranges with marked separation at higher stimulation intensities. Participants were retained when Sham and Active baseline MEP recruitment curves differed primarily in amplitude scale or estimated threshold but preserved a broadly similar monotonic recruitment profile. This led to the exclusion of four participants. Additionally, one participant had to be excluded because of an error in the patching procedure on one of the two experimental days. Data from a final set of n = 22 participants (8F, 14M, mean age = 29.05, SD = 5.1) were analyzed. Ethics The study was conducted following the guidelines of the Declaration of Helsinki (DoH-Oct2008) and received ethical approval from the Comité de Protection des Personnes Île-de-France III (2021-A03204-37). Each participant signed a written informed consent form and received financial compensation at a rate of 20€ per hour of the experiment. Apparatus & stimuli Magnetic resonance imaging Each participant underwent an magnetic resonance imaging (MRI) recording session for acquisition of a high resolution structural T1 3D sequence (T1-MP2RAGE images on a Siemens Magnetom Prisma Fit 3T scanner, Flip angle = 4, TR = 5000 ms, TE = 3.24 ms, slice thickness = 1 mm, isovoxel), which was employed in a frameless MRI-based stereotaxic neuronavigation system to localize and mark the stimulation area in M1 and to warrant stable coil positioning and targeting throughout and across experimental sessions. Psychophysics Participants completed the binocular rivalry task in a quiet and dimly lit room. Visual stimuli were generated in MATLAB (R2023b, The MathWorks Inc., Natick, MA) using Psychtoolbox-3 (Brainard 1997). The software ran on a PC (Dell Precision 3260, intelR Core i7) equipped with an NVIDIA T1000 graphics card (NVIDIA Corporation, Santa Clara, California, USA). An LCD monitor (Dell P2714Hc 1920 × 1080 pixels, 76 Hz refresh rate) was used for stimuli presentation and responses were recorded through a keyboard. Participants viewed the visual stimuli through a portable mirror stereoscope (ScreenScope, Berezin, Mission Viejo, California, USA) mounted to a custom-made forehead and chin rest, stabilizing their head at 36 cm from the screen. Two sinusoidal gratings with orthogonal orientations (45° clockwise or counterclockwise, size: 2°, spatial frequency: 2 cpd, contrast: 50%) were presented on a uniform gray background (luminance: 75.0 cd/m2, CIE x = 0.366, y = 0.383). Stimuli were displayed in central vision and included a central white fixation point and a common squared white frame to facilitate dichoptic fusion (Fig. 1a). Experimental methods. a) Binocular rivalry paradigm. Participants are presented with orthogonally oriented gratings on each eye through a mirror stereoscope. Participants are instructed to report their perception as their visual perception alternates between the two images, using the three arrow keys (right: clockwise, left: counterclockwise, down: mixed). b) Illustration of a primary motor cortex (M1) motor hotspot used as TMS target site highlighted from three different points of view of a representative participant. c) I/O recruitment curve procedure. MEPs were recorded from the FDI muscle at the base of the index finger, in response to single TMS pulses delivered to this muscle representation in the contralateral primary motor cortex. The peak-to-peak amplitude of the MEP is extracted and averaged across trials at the same TMS intensity and plotted across increasing intensity levels. Note that as expected, increases in TMS intensity (incremented as percentages relative to the resting motor threshold) induce an increase in the MEP amplitude. This relationship is compiled in an I/O recruitment curve plot and estimated by the area under the recruitment curve, before and after rTMS. For MD, custom-made translucent plastic eye-patches were used. Complying with previous human MD studies (Lunghi et al. 2011, 2013; Castaldi et al. 2020), these patches prevent pattern vision but let light reach the retina. Electromyography (EMG) Surface electromyography (EMG) was collected from the first dorsal interosseous (FDI) muscle, with an active electrode placed over the belly of the FDI muscle, a reference electrode on the distal tendinous junction of the index finger, near the base of the distal phalanx and a ground electrode was placed over the ipsilateral wrist, at the prominence of the radial bone, in contact with the carpal region. EMG signals were amplified using a Digitimer D360 amplifier (gain ×1000) and recorded with specialized software (Signal, version 5.0; Cambridge Electronic Design, UK). MEPs in response to single TMS pulses were collected and stored for offline analysis. The EMG signal was bandpass filtered from 0.02 to 2 kHz and digitized at 5 kHz sampling rate. Transcranial magnetic stimulation Single pulses and repetitive TMS patterns were delivered using a Magstim Rapid2 biphasic stimulator (Magstim Company, Whitland, UK) attached to a 70 mm figure-of-eight coil with integrated cooling system. Single-pulse TMS for the I/O recruitment curves (see upcoming section for further details on such TMS dose-MEP response recordings) was delivered using a Magstim BiStim2 monophasic stimulator (Magstim Company, Whitland, UK), also attached to a 70 mm figure-of-eight coil. An MRI-based neuronavigation system (Brainsight, Rogue Research, Montréal, Canada) was employed to track TMS coil position during the whole session. During TMS delivery, EEG activity was continuously monitored with a 62-electrode TMS-compatible EEG system (Actichamp, 5 kHz, Brain Products GmbH, Gilching, GR) providing additional safety monitoring and allowing ulterior exploratory analyses out of the scope of the current study. Procedures Participants were briefed about the experiment prior to their arrival at the lab. Once they had signed their consent forms, they completed a screening session to be admitted to the study. For participants who fulfilled the inclusion and exclusion criteria, three more sessions followed: an MRI acquisition and two experimental sessions counterbalanced across participants, one delivering Active and the other Sham 10 Hz stimulation during MD. The two experimental sessions were separated of at least 3 days (mean intersession time = 10.14 days, SD = 14.7, max = 66) to avoid unlikely TMS carryover effects across the two visits. Because of evidence suggesting potential interactions of visual homeostatic plasticity with energy metabolism (Lunghi et al. 2019) and circadian rhythms (Menicucci et al. 2022), each TMS condition was carried out on each participant at approximately the same time of the day and in a similar metabolic state (full stomach). Screening session Participants who declared interest and self-evaluated eligibility were given an appointment for a screening session which started with repeated briefing and the signing of informed consent form. Their adherence to the inclusion and exclusion criteria was verified by the experimenters. Ad hoc safety questionnaires were used to ensure MRI and TMS involved no risk for participants (Shellock and Spinazzi 2008; Rossi et al. 2009). After these steps, visual acuity of the participants was individually tested with “Early Treatment of Diabetic Retinopathy Study” (ETDRS) charts. Participants with more than 0.1 LogMar acuity for at least one eye were excluded from the study. After the visual acuity test, sighting eye dominance was determined with the “Porta test” (Lederer 1961). At the end of these tests, participants completed a training (3 min) and a full block (6 min) of the binocular rivalry task explained in the following section. Those who had abnormal binocular fusion (eg were unable to merge the two images projected to the two eyes into a single perception or experienced more than 30% of mixed percepts) were excluded from the rest of the study. We also excluded participants with strong dominance of one eye, defined as a baseline ocular dominance > 1.7 (see Equation 2). Psychophysics Ocular dominance was assessed using binocular rivalry (Levelt 1966; Alais and Blake 2005), and visual plasticity was defined as a shift in ocular dominance following short-term (90 min) MD. To this end, participants viewed stimuli through a mirror stereoscope and indicated their percept using arrow keys: left for counterclockwise grating, right for clockwise grating, and down for mixed percepts (Fig. 1a). On each of the two experimental sessions, participants started with a training session (3 min) of binocular rivalry, followed by a baseline block. Each rivalry block lasted 8 min, composed of two 3-min trials separated by a pause. Each 3-min trial included two 90-s sub-trials and a 3-s break. Grating positions were alternated between sub-trials to avoid adaptation, and initial positions were counterbalanced across participants. Participants wore an eye-patch over their dominant eye for 90 min, placed after the baseline binocular rivalry block (Fig. 2). Dominance was defined by the eye that dominated perception more frequently during the baseline block of binocular rivalry. When baseline ocular dominance remained unclear according to this criterion, the “Porta test” was employed (Lederer 1961). During MD and as 10 Hz repetitive TMS was delivered, participants watched a TV show with subtitles on computer screen placed at arm’s reach (or ~57 cm from their eyes canthi). After MD, they completed two additional rivalry blocks—one immediately after patch removal and another 30 min later. The deprivation index (see Equation 1) was computed and quantified at each post-block as a proxy of visual plasticity. Experimental pipeline. Participants performed a binocular rivalry task, in which they were requested to indicate the orientation of the grating they perceived, before and after 90 min of monocular deprivation. Thirty minutes into the deprivation period, Active or Sham 10 Hz rTMS was delivered to the primary motor cortex for 28 min. At the start of the monocular deprivation, and immediately after the end of the rTMS, the I/O curve was recorded. Transcranial magnetic stimulation Participants were seated in a comfortable chair and instructed to keep both arms and hands in a relaxed position throughout the experiment. An MRI based frameless stereotaxic neuronavigation system (Brainsight, Rogue Research, Massachusetts, USA) was employed to avoid TMS location shifts from the cortical site designated as the M1 target, while ensuring reliable targeting with a maximal ±2 mm variability in all four directions around individually established target coordinates. As the effect of MD has not been shown to be hemisphere-specific (Lunghi et al. 2015b; Binda et al. 2018), participants received rTMS either on their left or right hemisphere in counterbalanced order across participants. The TMS coil was held on the individually established target position tangentially to the surface of the participant’s head, with the handle oriented caudal to rostral and medial to lateral (~45° from the head longitudinal midline, hence perpendicular to the central sulcus) to induce a posterior-to-anterior (PA) current, following the conventions in brain stimulation (Mills et al. 1992; Brasil-Neto et al. 1993; Rossi et al. 2021). The target stimulation site always was the representation of the FDI muscle in M1. Motor hotspot At the onset of each experimental session, following a baseline binocular rivalry recording block, we identified and labeled on the participant’s MRI the localization of their individual motor hotspot for the FDI muscle. To successfully complete this M1 hotspot search operation (which took on average 30 min), single TMS pulses were delivered along and in the vicinity of the precentral gyrus as visualized in the neuronavigation system, aiming to observe a twitch of the contralateral index finger. TMS pulse intensities were first adjusted in real time to elicit consistently visible FDI muscle twitches. EMG recordings from the FDI muscle were employed to further verify the scalp and projected cortical site ultimately selected as M1 hotspot by consistently identifying the location eliciting motion activity as specific as possible on the index finger. This position was then labeled on the neuronavigation system and employed for all further TMS operations (Fig. 1b). Resting motor threshold After determining the FDI motor hotspot, we measured the individual resting motor threshold defined as the minimum intensity of stimulation able to generate an MEP of at least 50 μV peak-to-peak amplitude in at least five out of ten consecutive trials (Rossini et al. 1994), with single TMS pulses delivered at randomly variable intervals between 4 to 7-s of duration to avoid cumulative inter-pulse effects or interference of MEP amplitude by participant’s anticipation. We then estimated the TMS pulse intensity steps necessary to compile a complete I/O curve and to dose the 10 Hz rTMS patterns, in both cases, relative to each individual resting motor threshold. Since single-pulses for the I/O curve, and 10 Hz trains for rTMS were delivered with different stimulation equipment, the resting motor threshold was determined separately for each device. The MD period started once the resting motor threshold evaluation was complete. I/O curve An I/O (aka recruitment or dose–response) curve was compiled for each participant and session (Fig. 1c) before (baseline-I/O) and after (post-I/O) rTMS application, during MD (Fig. 2). A total of 60 pulses, 10 pulses each at intensities of 80%, 90%, 100%, 110%, 130%, and 150% relative to each subject’s individual resting motor threshold, were delivered in progressive order. An inter-stimulus interval between 4 and 7 s was inserted to avoid carryover effects across successive pulses (Chen et al. 1997). Changes in cortical excitability of each participant were evaluated by recording the peak-to-peak amplitude of MEPs at different intensity levels. These data were used to compile I/O recruitment curves collected immediately prior and following rTMS at the same stimulation intensities. Repetitive transcranial magnetic stimulation All rTMS procedures were identical for Active and Sham 10 Hz rTMS conditions, performed in two separate and independent sessions (Fig. 2). “Sham” 10 Hz rTMS was delivered with a standard rTMS coil with its lateral edge positioned tangential to the scalp and its main surface at a 90° angle relative to the target area, hence directing the TMS field away from the scalp. The order of Sham and Active 10 Hz rTMS stimulation sessions was counterbalanced across participants to avoid order bias effects. Repetitive TMS pulses were applied for a duration of 28 min, starting after the first 30 min of MD. This timing was chosen to maximize the temporal overlap of rTMS effects and MD effects: MD effects increase with longer durations of deprivation (Min et al. 2022; Prosper et al. 2023), 28 min long 10 Hz rTMS patterns are expected to produce up to 30 min aftereffects (Van Der Werf and Paus 2006; Thut and Pascual-Leone 2009). This timeframe left at least a 50 min epoch of overlap of visual and motor cortex plasticity manipulations. High-frequency rTMS was delivered at 10 Hz for a total duration of 28 min (100 ms inter-pulse interval, 40 pulse trains, repeated 40 times, for a total of 1600 pulses, with 40 s intertrain intervals). Stimulation intensity was set at 90% of the individual resting motor threshold, although in 8 participants the intensity was further reduced (up to 17% of the total device output, on average by 11.5%), to avoid sustained scalp discomfort. The mean stimulation intensity for the cohort of participants that completed the study (n = 22) was 83 ± 9% of individual resting motor threshold and 65 ± 11% of maximum stimulator output. The higher than typical range of absolute resting motor thresholds in this study was due to the EEG cap worn by the participants during the sessions, which added 5 mm distance between the coil surface and the scalp. The rTMS parameters for excitatory stimulation employed were chosen following previous studies (Pascual-Leone et al. 1994; Maeda et al. 2000b) and generously respected at all times the safety limitations established by the most current version of the international guidelines regulating TMS uses (Rossi et al. 2021). To protect the hearing of the participants, to avoid transient post-rTMS hypoacusia, and also to suppress the impact of TMS clicking noise, participants were exposed to masking white noise played through foam earplugs with an MRI/TMS compatible earphones system (ER-3C insert earphones, ETYMOTIC Research, Chicago, USA) (ter Braack et al. 2015). White noise level was individually adjusted by gradually increasing the volume until the participant stopped hearing TMS coil clicking or reached their threshold of auditory discomfort. At the end of the Active or Sham 10 Hz rTMS block, participants pursued MD for approximately 30 additional minutes during which the post rTMS I/O curve was recorded (Fig. 2). At the end of each experimental session, participants filled a standardized TMS tolerability questionnaire. Data analysis Psychophysics The recorded binocular rivalry perceptual durations, defined as “phase durations,” were first processed with MATLAB and then on R for statistical analyses. Phase durations < 0.25 s were discarded to eliminate keypress errors. The mean phase duration (MPD) of each eye (deprived or non-deprived) for a given block of binocular rivalry was calculated with the geometric mean to account for the non-normal distribution nature of binocular rivalry data (Levelt 1967). To compare the change in ocular dominance following MD in the Active and Sham 10 Hz rTMS conditions, we calculated a deprivation index (Lunghi et al. 2015b; Binda and Lunghi 2017; Binda et al. 2018; Menicucci et al. 2022; Acquafredda and Binda 2024): where Baseline MPD indicates the mean phase duration at baseline, and MPD refers to the mean phase duration for the block concerned. deprivation index values around 1 indicate, therefore, balanced ocular dominance, while deprivation index < 1 indicates predominance of the deprived eye, and deprivation index > 1 predominance of the non-deprived eye. We then compared the deprivation index across two rTMS conditions (Active vs Sham) using a 2 (blocks, excluding baseline) ×2 (conditions) repeated-measures ANOVA. To compare the Active and Sham 10 Hz rTMS conditions in relation to the motor stimulation effect, the deprivation index was log10-transformed and the Sham rTMS value subtracted from the Active 10 Hz rTMS value. The visual plasticity potential was approximated as the log10-transformed deprivation index in the Sham 10 Hz rTMS condition. The baseline ocular dominance was calculated as follows: The correlations were computed with Spearman’s rank correlation and further validated with permutation and bootstrap-based estimates using the robust correlation package on MATLAB (Pernet et al. 2013). To account for multiple correlation tests, the corresponding P-values were additionally adjusted using Holm correction. I/O curve MEP recordings were first analyzed on Signal 5.0 (CED, Cambridge Electronic Design Ltd, UK). EMG traces from the baseline-I/O and post-I/O were first visually inspected. The frames lacking a single pulse TMS artifact, including a movement artifact, or unusual noise were excluded from further analyses. The remaining frames were binned into intensities with ~ 10 trials in each pre-set intensity level per participant. Note that the individually adjusted single pulse TMS stimulation intensity, together with the limited maximum output provided by the TMS device led to n = 5 subjects not being stimulated at 150% of their individual resting motor thresholds (which would have required higher than 100% of maximum stimulator output intensity). Additionally, n = 2 other participants were not stimulated at 150% intensity because they did not tolerate well the highest intensity levels recorded to complete the full recruitment curve. Statistical analyses were performed on trial-by-trial data to avoid the influence of above-mentioned missing values, particularly at the highest TMS intensity. Optimal MEP time windows to exclude single-pulse TMS artifacts and include the full range of peak-to-peak signal were defined for each participant and rTMS condition. For each trial, we calculated the peak-to-peak amplitude (in mV) of the MEP following the TMS pulse artifact. The extracted MEP parameters were transferred to R for further analysis. For every participant, the trial-by-trial plot of peak-to-peak amplitudes was drawn for each condition (Active or Sham 10 Hz rTMS) and time point (baseline or post-rTMS) separately and the area under the recruitment curve (AUC) was calculated using trapezoidal integration. These AUC values were compared in a 2 (condition) ×2 (time) repeated-measures ANOVA. To be able to associate the effects of rTMS on the primary M1 motor cortex (motor plasticity) and the impact of MD (visual plasticity), we also calculated for each rTMS condition a motor excitability index (comparable to the deprivation index used for visual plasticity): A motor excitability index > 1 indicates a shift toward an excitation (excitatory change) in the corticospinal excitability (ie higher peak-to-peak amplitudes at analogous stimulation intensities) whereas a motor excitability index < 1 denotes a shift toward inhibition (inhibitory change) following Active10 Hz rTMS stimulation on M1. To explore potential correlations between the motor excitability index and the visual deprivation index, values were log10 transformed and Sham 10 Hz rTMS values were subtracted from the Active 10 Hz rTMS values. Additionally, the motor plasticity potential was approximated by the log10-transformed motor excitability index in the Active condition. As a complementary analysis, we also fitted a linear mixed-effects model with deprivation index as the dependent variable, motor excitability, stimulation condition, and their interaction as fixed effects, and participant as a random intercept, and compared models with and without the interaction term using a likelihood-ratio test. The full model was built as follows: Deprivation Index ~ Motor Excitation Index * Condition + (1|Participant). Results We investigated the impact of high frequency stimulation of the primary motor cortex (M1) on short-term visual homeostatic plasticity in a group of healthy adults (n = 22, Fig. 2). The lasting after-effects on motor cortico-spinal excitability were induced by rTMS and measured with input/output curve recordings of associated motor evoked potentials (MEPs) elicited with single pulses at increasing levels of intensity prior to and immediately following the end of the 28-min-long Sham or Active 10 Hz rTMS (Fig. 1c). Visual homeostatic plasticity was induced by short-term (90 min) MD and measured as the MD-induced shift in ocular dominance during binocular rivalry (Fig. 1a). To assess the test–retest reliability of our dependent variables, we compared the area under the MEP baseline-I/O recruitment curve (AUC) and the baseline ocular dominance (see Eq. 2) in the two experimental conditions (Active and Sham 10 Hz rTMS). We found strong correlations across participants for the two baseline measurements in both the baseline-I/O motor recruitment curve (rho = 0.72, P = 0.00014, Holm-adjusted P = 0.0254; validated by robust correlation estimates Spearman’s r = 0.558442, P = 0.007, CI = [0.0849003 0.873935], Fig. 3a) and the baseline ocular dominance (rho = 0.88, P = 0.00000267, Holm-adjusted P = 0.0000133; robust correlation Spearman’s r = 0.882552, P < 0.0001, CI = [0.725714 0.939394], Fig. 3b), indicating reliable estimations of both motor cortex excitability and ocular dominance across separate days. Baseline correlations. a) Correlation between the area under the curve of the I/O recruitment curve measured at baseline (baseline-I/O) in the Active vs. Sham 10 Hz rTMS conditions (performed on two different sessions at least 72 h apart in counterbalanced order). Each symbol is a single participant. Regression line drawn in blue. Dashed gray lines indicate x = y. b) Same as (a) but for ocular dominance. Having assessed the reliability of our measurements, we then confirmed the efficacy of each experimental manipulation (10 Hz rTMS and short-term MD), in inducing motor and visual plasticity according to the expected main effects (a group mean increase in motor cortico-spinal excitability measured by the increase of the area under the curve of the I/O MEP amplitude recruitment curve, and ocular dominance shift toward the deprived eye). Hence, in line with well-established observations for conventional rTMS patterns, Active high frequency stimulation at 10 Hz increased the amplitude of the individual average MEPs and the area under I/O response at the group level (Fig. 4a). A 2 (condition) × 2(time) repeated measures ANOVA revealed no significant effect of time (F(1,21) = 3.02, P = 0.09, η2p = 0.1652, or condition (F(1,21) = 1.48, P = 0.23, η2p = 0.0554) but detected a significant interaction between the two (F(1,21) = 5.80, P = 0.025, η2p = 0.2141). Participant-level plot revealed that most participants showed changes in an excitatory direction following Active 10 Hz rTMS. Nonetheless, compared to natural excitability fluctuations gauged by the Sham 10 Hz rTMS condition, Active 10 Hz rTMS impact was marked by a considerable level of inter-individual variability in the strength and the direction of the net modulatory effects, (Fig. 5a), all well-documented in previous literature (Pascual-Leone et al. 1994; Maeda et al. 2000a, 2002; Nettekoven et al. 2015; Corp et al. 2020; Goldsworthy et al. 2021). Group average results. a) Average I/O curves obtained in the Sham (red) and Active (blue) 10 Hz rTMS conditions. Baseline curves are traced with dashed lines (baseline-I/O), and solid lines are post-rTMS measurements (post-I/O). b) Group average deprivation index (DI) for the Sham (red symbols) and the Active (blue symbols) 10 Hz rTMS conditions. Error bars represent 1 ± S.E.M. Individual results. a) Scatterplot of the difference in post-I/O and baseline-I/O area under the curve measurements (excitation index) for each participant in Sham and the Active 10 Hz rTMS conditions. Each symbol is a single participant. Dashed gray lines indicate x = y. b) Scatterplot of participant deprivation index measured immediately after deprivation (4 min) for the Sham and the Active 10 Hz rTMS conditions. Each symbol is a single participant. Dashed gray lines indicate x = y. In line with previous reports, 90 min of MD induced a significant shift of ocular dominance in favor of the deprived eye, as indicated by a change in the deprivation index (Eq.1) for the Active (t(21) = −7.5223, P = 2.178e-07, d = −1.60) as well as the Sham rTMS conditions (t(21) = −7.2254, P = 4.048e-07, d = −1.54, Fig. 4b). To investigate whether motor cortex rTMS stimulation impacted visual homeostatic plasticity, we first compared the effect of short-term MD in the Sham and Active 10 Hz stimulation conditions. We found that at the group level, the ocular dominance shift induced by MD was not impacted by the presence of Active motor cortex stimulation (Fig. 4b). A 2 (rTMS condition) × 2 (block) repeated-measures ANOVA of the deprivation index revealed a significant effect of block (F(1,21) = 26.4, P = 0.00004, η2p = 0.557), indicating a trend for return to baseline values, in line with previous literature. However, no significant effect of rTMS condition (F(1,21) = 0.16, P = 0.68, η2p = 0.0077) or an interaction between the former and block (F(1,21) = 0.001, P = 0.96, η2p = 0.00007) was found (Fig. 4b). Participant-level plot revealed variability in the strength of the effect across conditions (Fig. 5b). Factors such as handedness, stimulated hemisphere, and eye-patch positioning did not systematically impact these outcomes (see Fig. S1 for details). This lack of a main effect of TMS was not surprising, given the well-documented inter-individual variability in rTMS effects (Pascual-Leone et al. 1994; Maeda et al. 2000a, 2002; Nettekoven et al. 2015; Corp et al. 2020; Goldsworthy et al. 2021). Our data also revealed that some participants showed decreased corticospinal excitability after Active 10 Hz rTMS, although most showed an increase (Fig. 5a). To consider these crucial individual differences and to be able to appreciate the multifaceted impact of M1 stimulation on visual plasticity, we computed the interaction of our two manipulations at the individual participant level. We proceeded to investigate the correlation between the modulation in the cortico-spinal excitability induced by rTMS (see Fig. S2 for MEP profiles) and the difference in the ocular dominance shift induced after the 10 Hz rTMS pattern. To this aim, we computed the correlation between the deprivation index measured immediately after MD and the motor excitation index, subtracting the Sham 10 Hz rTMS values from the Active 10 Hz rTMS values. This subtraction allowed us to estimate how Active rTMS modulated the MD-induced ocular dominance shift beyond MD itself, while also accounting for nonspecific changes of corticospinal excitability across time. We found a strong correlation between the two (rho = 0.54, P = 0.0101, Holm-adjusted P = 0.0254, Fig. 6a). Further evaluation with a permutation test proved this correlation to be robust (Spearman’s r = 0.54, P = 0.009, CI = [0.108447 0.802385]). Note that as indicated previously, a deprivation index < 1 was expected after MD, and consequently, a higher deprivation index denotes a smaller ocular dominance shift following MD. On the other hand, for the motor excitability index, values > 1 signal increased excitability following 10 Hz rTMS. Hence, the positive correlation here reported indicates that participants showing increased motor cortex excitability following Active 10 Hz rTMS (compared to the Sham 10 Hz rTMS condition) also showed decreased visual plasticity in the former (compared to the latter). As a complementary analysis, the linear mixed-effects model did not reveal a significant interaction between motor excitability and stimulation condition (β = −0.386, SE = 0.299, t = −1.29, P = 0.204), and model comparison indicated that including the interaction term did not significantly improve model fit (χ2(1) = 1.58, P = 0.209). Visual and motor correlations. a) Correlation between the rTMS effects on the deprivation index (Active – Sham 10 Hz rTMS) and excitation index (Active – Sham 10 Hz rTMS). Each symbol is a single participant. Regression line drawn in blue. b) Correlation between the sole plasticity indices: Deprivation index in the sham condition for visual plasticity and excitation index in the Active 10 Hz condition for motor plasticity. Each symbol is a single participant. Note that a deprivation index < 1 is a proxy of plasticity, with a lower score denoting higher levels of plasticity. In contrast, an excitation index > 1 is used as a proxy of excitability changes, meaning higher levels of excitation, whereas a score < 1 is signals inhibition. Regression line drawn in blue. To further investigate the link between the plastic potential on M1 modulated by 10 Hz rTMS and early visual areas involved in ocular dominance plasticity, we analyzed the effects of MD and transcranial stimulation separately. Our previous study (Sarı and Lunghi 2023) provided evidence that motor plasticity can interfere with visual plasticity whereas visual plasticity does not impact motor plasticity. Additionally, we did not find a correlation between the motor and visual plasticity scores when induced in isolation. Motivated by these findings, we performed an exploratory analysis testing whether visual and motor plasticity were related in the present dataset. We correlated the deprivation index in the Sham rTMS condition, reflecting visual plasticity without active M1 stimulation, with the motor excitability index in the Active rTMS condition, taken as an approximate index of each participant’s rTMS-induced motor plasticity. We acknowledge that this motor index was obtained in presence of MD, however, based on our previous observation that visual plasticity did not measurably influence motor plasticity, we considered it a reasonable exploratory proxy. We found a strong negative correlation between the two indices (rho = −0.57, P = 0.00635, Holm-adjusted P = 0.0254, Fig. 6b), validated further with robust correlation estimates (Spearman’s r = −0.57, P = 0.00552541, CI = [−0.835227–0.143347]). As a smaller deprivation index indicates a larger ocular dominance shift, this negative correlation points toward a positive relationship between the plastic potential operated by 10 Hz rTMS on M1 and that occurring in early visual cortex in adult humans induced by MD. Discussion To the best of our knowledge, our study provides for the first time evidence that directly modulating motor excitability in adult humans may influence the magnitude of visual plasticity phenomena. Specifically, altering cortico-spinal excitability with rTMS applied to M1 modulated the magnitude of ocular dominance plasticity in the opposite direction. Our data also revealed a positive correlation between the plastic potentials of these two cortical regions, estimated by responses to short-term MD (resulting in ocular dominance change) and rTMS (inducing motor cortico-spinal excitability shifts), respectively (Fig. 6b). Taken together, our results challenge the notion of brain plasticity as a strictly modular phenomenon limited by the anatomical boundaries of regions associated with the same type of functions. They support instead a network perspective, by which plasticity events are regulated across interconnected systems, well beyond the boundaries set by cortices associated with different sensory modalities. We analyzed the local plastic effects of rTMS by assessing changes in MEP amplitudes across different single-pulse TMS intensities, and examined their relationship with visual cortex plasticity, as measured by ocular dominance shifts following short-term MD. We found a negative correlation between the two; Specifically, participants with a greater shift of M1 excitability following 28 min of Active rTMS exhibited smaller ocular dominance shifts after MD compared to Sham stimulation. Our data suggest that stronger shifts of M1 excitability post rTMS were associated with reduced visual cortex plasticity following MD, suggesting that increased M1 excitability may compete with the visual cortex resources necessary for plastic expression, limiting the extent of visual plasticity. We complemented the correlation analysis with a mixed-effects model including stimulation condition, motor excitability, and their interaction. This analysis did not reveal a significant motor excitability × condition interaction, indicating that the condition-specific association should be interpreted cautiously and motivating replication in a larger sample. Interestingly, we also observed a positive correlation between the intrinsic plastic potentials of these two types of cortices: individuals with higher motor plasticity following Active rTMS tended to exhibit higher visual plasticity after Sham stimulation. Taken together, these results and the method employed to generate them suggest a likely direct association between visual and motor cortex plasticity in adult humans. Pioneering studies investigating cortical excitability as measured by TMS-evoked phosphene or TMS-determined M1 motor thresholds found mixed results ranging from null (Boroojerdi et al. 2002) to marginal correlations (Deblieck et al. 2008) between these two measures. The outcomes of a recent meta-analysis on this same matter supported a correlation between these two types of measurements (Phylactou et al. 2024). Unfortunately, none of these reports, indexing associated levels of motor and visual excitability, enabled a direct evaluation of the plastic potentials in M1 and V1. Another study reported a correlation between the potentiation of visual evoked potentials and M1 paired-associate stimulation, suggesting a link between LTP-like processes in the motor and visual systems (Klöppel et al. 2015). Taken together, these accounts draw support for the hypothesis that neuroplasticity is a trait-like aspect of brain systems, expressed under normal circumstances similarly in both motor and visual cortices. Moreover, this view entertains the possibility that when these phenomena are engaged in motor and visual systems simultaneously, plasticity throughout these two types of networks may enter a competitive regime, limiting the extent of the plastic reorganization or changing some of its features. Interestingly, this framework is aligned with our current findings and also prior reports by our team showing diminished ocular dominance plasticity in the presence of motor learning (Sarı and Lunghi 2023) and behavioral accounts that motor learning influences visual performance, whereas conversely visual learning did not affect motor performance (Cates and Gordon 2024). In this context, a plausible framework to interpret past and current outcomes is the existence of rivalrous interactions between systems and mechanisms subtending motor and visual plasticity, aligned with the notion of global homeostatic regulation of brain plasticity, whereby local plastic changes in the healthy brain are strongly constrained by overarching mechanisms aiming to preserve the stability of the widespread networks connecting brain systems beyond modalities and functions. Extending this view, while classical homeostatic plasticity has shown to locally balance E/I within a circuit (Turrigiano 1999), here we propose that the brain also enforces system-wide equilibrium. When one network undergoes strong potentiation, others may transiently reduce their capacity for change in E/I, maintaining overall excitability within a safe and metabolically sustainable range (Marder and Goaillard 2006; Abraham 2008). Indeed, studies in mice have shown that a stroke in M1 or S1 abolishes ocular dominance plasticity completely (Greifzu et al. 2011; Pielecka-Fortuna et al. 2015), suggesting that the lesion-induced changes in E/I in these regions generate a response that limits visual plasticity, despite its distance from the impacted area. This framework could provide a plausible explanation for our findings and explain why in a scenario of rTMS-driven increases of cortico-spinal M1 excitability, compensations arise and limit concurrent plastic reorganization in the visual cortex. The relationship observed here likely emerges from a combination of local and global mechanisms. Both tDCS (Nitsche et al. 2005; Stagg et al. 2009a; Stagg et al. 2011) and TMS stimulation with theta bursts (Stagg et al. 2009b) lead to changes in local GABA and Glutamate concentrations when applied to the motor cortex. While direct evidence on the relationship between conventional rTMS and GABA/Glutamate concentration with NMR spectroscopy in healthy humans remains still scarce or controversial, previous studies [for reviews see (Fitzgerald et al. 2006; Stagg et al. 2010; Stagg 2014; Valero-Cabré et al. 2017; Chhabra et al. 2021; Cuypers and Marsman 2021; Jannati et al. 2023)] warrant the hypothesis that high frequency rTMS over M1 may modulate local GABA and/or Glutamate concentration, thereby shifting the E/I balance upwards. These effects are not strictly local: stimulation can propagate to anatomically and functionally interconnected sites within extended networks, including parietal and prefrontal regions, or homotopic loci of the contralateral hemisphere (Stagg et al. 2010; Valero-Cabré et al. 2017), changing the E/I balance indirectly. Notably, classical feline studies combining low and high-frequency rTMS with 14C-2Deoxyglucose ultra-high-resolution imaging show that stimulation effects propagate trans-synaptically to anatomically connected regions (Valero-Cabré et al. 2005, 2007). These remote effects scale with the strength of underlying white-matter projections, demonstrating that rTMS influences distributed cortical networks (Valero-Cabré et al. 2007). Such network-level effects could divert shared neuromodulatory resources away from V1, limiting its plastic capacity. In this sense, adult plasticity can be viewed as a globally regulated resource rather than an independent property of isolated regions. Such regulation may occur through shared neuromodulatory systems, or long-range projections, under strict metabolic constraints, ensuring that adaptive flexibility does not come at the cost of network instability. It is important to note that, at the group level, rTMS on M1 failed to produce a consistent change in ocular dominance plasticity. This is not surprising, given the well-documented variability in individual responses to rTMS (Pascual-Leone et al. 1994; Maeda et al. 2000a, 2002; Nettekoven et al. 2015; Corp et al. 2020; Goldsworthy et al. 2021), which requires an individual differences analysis approach in order to appreciate the impact of M1 stimulation on visual cortical plasticity. While the 10 Hz protocol is generally excitatory, high frequency patterns up to 20 Hz have shown inhibitory modulations in some individuals (Gangitano et al. 2002), and the magnitude of the E/I shift varies widely (Maeda et al. 2000a; Corp et al. 2020). Although in our study TMS stimulation intensity was adjusted to each participant’s resting motor threshold, some required lower intensities due to discomfort, whereas even at 90% resting motor threshold, some individuals showed weaker excitatory responses than others. These factors, combined with a modest sample size, might have reduced the power to detect smaller size group-level effects. Accordingly, the absence of a significant Active versus Sham rTMS difference at the group level should be interpreted cautiously, as it may reflect limited power rather than a true absence of an effect. At the same time, sample sizes in a similar range have been sufficient in previous studies to detect robust effects of short-term MD and conventional rTMS, suggesting that any cross-modal group-level effect in the present study may be relatively small and variable across participants. Importantly, the complementary mixed-effects analysis did not reveal a significant motor excitability × rTMS condition interaction. Thus, while the difference-score correlation supports an association between individual rTMS-induced changes in motor excitability and changes in ocular dominance plasticity, this condition-specific relationship should be considered preliminary and should be tested in larger samples. Our protocol was designed to maximize temporal overlap between MD and rTMS. Stimulation duration was 28 min, a duration established as safe and effective (Maeda et al. 2000b; Gangitano et al. 2002; Houdayer et al. 2008). Because single-session rTMS aftereffects could last up to 30 min (Van Der Werf and Paus 2006; Thut and Pascual-Leone 2009), this resulted in a large overlap with the 90-min MD period. However, note that to limit interference with ongoing M1 cortical excitability/plasticity changes using additional pulses or trains, we decided not to measure the decay of rTMS effects over time. By the end of MD, the excitatory effect may have partially or completely diminished in some participants, raising the mechanistically important question of whether rTMS primarily prevented the build-up of ocular dominance plasticity or disrupted plasticity that had already developed. Our results nevertheless support the idea that rTMS disrupted the progressive build-up of the ocular dominance shift during deprivation. Shorter periods of MD produce smaller ocular-dominance shifts, indicating that the effect develops progressively over time (Min et al. 2022; Prosper et al. 2023). Thus, even if the effect of rTMS decayed before the end of MD, its influence during the early part of deprivation may have been sufficient to reduce the final ocular-dominance shift. The present design does not allow us to demonstrate this mechanism directly, and other mechanisms may also have contributed, including a disruption of plasticity that had already developed. Finding a design allowing the tracking of real-time changes in excitability during and after stimulation, interfering the least with ongoing plastic processes will be however crucial to disentangle these possibilities. In our previous study (Sarı and Lunghi 2023), motor learning not only altered the E/I balance in M1 but also induced systematic and long-lasting plastic changes, in addition to the involvement of the prefrontal cortex and working memory-related processes. In contrast, rTMS-induced plasticity exhibits dynamics typically not resulting in lasting reorganization nor involving working-memory-related mechanisms. While the present results demonstrate that changes in M1 excitability are associated with reductions of visual plasticity, rTMS effects may have been too transient or unstructured to produce a consistent group-level outcome. Another possibility is that prefrontal involvement, as in the motor learning paradigm or working memory task employed in our previous study (Sarı and Lunghi 2023), imposes a stronger influence on visual plasticity. Future work could test this idea by stimulating prefrontal regions, in addition to investigating whether sequential or alternating protocols, for example, MD followed by motor stimulation versus simultaneous induction, produce different outcomes, helping to identify optimal strategies for enhancing plasticity. Our results have potential clinical implications for learning and rehabilitation. If motor and visual cortices compete for plastic resources, simultaneous interventions in both domains may be less effective than staggered training schedules. For example, intensive motor rehabilitation could temporarily reduce the efficacy of visual therapies for amblyopia or stroke-related visual deficits. More broadly, the principle of global homeostatic regulation on plasticity likely extends beyond motor and visual cortices. Any intervention requiring E/I regulations in distinct brain regions should therefore consider the global regulation of plasticity, especially in the context of training and rehabilitation. Conclusion In summary, our findings lend support to the hypothesis that the motor and visual cortices are components of a shared, interconnected plasticity system. While individuals with high intrinsic plastic potential show strong responses in both cortices, simultaneous engagement of the two limits the ability of the visual cortex to adapt and privileges changes in motor systems. These results call for a reconsideration of the model of neuroplasticity, highlighting the need for a whole-brain framework when studying adult plasticity or when designing interventions to maximize learning and recovery across sensory and motor systems. Supplementary Material Acknowledgments The authors would like to thank Agathe Mangialomini for her help in early data collection. The authors declare no competing financial interests. Contributor Information İzel D Sarı, Laboratoire des Systèmes Perceptifs, Département d’Études Cognitives, École Normale Supérieure, Paris Sciences et Lettres University, Centre National de la Recherche Scientifique, 29, rue d'Ulm, Paris 75005, France. Vasileia Papakosta, Laboratoire des Systèmes Perceptifs, Département d’Études Cognitives, École Normale Supérieure, Paris Sciences et Lettres University, Centre National de la Recherche Scientifique, 29, rue d'Ulm, Paris 75005, France. Xavier Corominas Teruel, Danish Research Centre for Magnetic Resonance, Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital Hvidovre, Kettegård Alle 36, 2650 Hvidovre, Denmark. Hanna Willis, Laboratoire des Systèmes Perceptifs, Département d’Études Cognitives, École Normale Supérieure, Paris Sciences et Lettres University, Centre National de la Recherche Scientifique, 29, rue d'Ulm, Paris 75005, France. Antoni Valero-Cabré, Sorbonne Université, Institut Du Cerveau - Paris Brain Institute - ICM, Inserm, CNRS, Causal Dynamics, Plasticity and Rehabilitation Group, Frontlab Team, APHP, Hôpital de La Pitié Salpêtrière, 47-83 boulevard de l'Hôpital, Paris 75013, France. Claudia Lunghi, Laboratoire des Systèmes Perceptifs, Département d’Études Cognitives, École Normale Supérieure, Paris Sciences et Lettres University, Centre National de la Recherche Scientifique, 29, rue d'Ulm, Paris 75005, France. Author contributions Izel D Sari (Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Supervision, Visualization, Writing—original draft, Writing—review & editing, Software), Vasileia Papakosta (Investigation, Project administration), Xavier Corominas Teurel (Investigation, Software, Writing—review & editing), Hanna Willis (Investigation, Software, Writing—review & editing), Antoni Valero-Cabré (Conceptualization, Investigation, Methodology, Supervision, Writing—review & editing), and Claudia Lunghi (Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Visualization, Writing—original draft, Writing—review & editing) Funding This work was supported by the European Research Council (ERC) Horizon 2020 Research and Innovation Program Grant No 948366-HOPLA and by French National Research Agency Grants ANR-19-CE28-0008-PlaStiC and ANR-17-EURE-0017-FrontCog. Conflicts of interest None declared. Data availability Data and codes described in this study are publicly available in the repository Zenodo. DOI: 10.5281/zenodo.22042699. References Associated Data Supplementary Materials Data Availability Statement Data and codes described in this study are publicly available in the repository Zenodo. DOI: 10.5281/zenodo.22042699.