This observational study (n=20) of experienced ayahuasca users linked the drug’s subjective effects with changes in blood alkaloids, metabolism and brain network connectivity. The strongest shared patterns involved altered default mode and attention networks alongside lipid-related metabolic signals, suggesting the ayahuasca experience reflects coordinated brain-body changes.
Ayahuasca is a psychoactive brew containing N,N-dimethyltryptamine (DMT) and β-carboline alkaloids that induces marked alterations in perception, emotion and self-referential processing. However, the multiscale biological organization linking peripheral metabolism, brain network dynamics, neurochemistry, and subjective experience in humans remains poorly understood. Here, we applied an integrative, within-subject, multiblock partial least squares framework to model coordinated changes across four complementary biological and phenotypic layers: plasma psychoactive alkaloids, targeted metabolomics, resting-state fMRI-derived functional connectomes, and multidimensional subjective experience assessed with the 5-Dimensional Altered States of Consciousness (5D-ASC) scale, in 20 experienced ceremonial ayahuasca users. Complementary ¹H-MRS data were used to examine associations between peripheral metabolism, posterior cingulate cortex neurochemistry, and default mode network (DMN)-related connectivity. Multilayer integration revealed that the experiential dimensions oceanic boundlessness, visionary restructuralization and auditory alterations covaried with circulating DMT and β-carbolines, alterations in lipid, amino acid and energy metabolisms and reconfiguration of dorsal attention- and DMN-related connectivity. Shared network features across experiential dimensions were most strongly associated with endocannabinoid-related N-acylethanolamines, acylglycerols, and ceramides, extending canonical serotonergic models toward downstream lipid-signalling and metabolic processes. Complementary rCCA analyses further showed structured covariation between peripheral metabolites, posterior cingulate cortex neurochemistry, and DMN-related connectivity. Together, these findings indicate that psychedelic states reflect coordinated, system-level interactions between peripheral metabolism and functional brain networks rather than isolated neurochemical or neural events. Framed within a brain-body integromics perspective, this work provides translationally relevant insight into metabolic pathways that may modulate brain function and subjective response, with potential implications for neuropsychiatric and pharmacometabolic research.
Papers cited by this study that are also in Blossom
Riba, J., Valle, M., Urbano, G. et al. · Journal of Pharmacology and Experimental Therapeutics (2003)
Äbelö, A., Smallridge, J. W., Von Rotz, R. et al. · Biomedicine & Pharmacotherapy (2025)
Madrid-Gambin, F., Gomez-Gomez, A., Busquets-Garcia, A. et al. · Biomedicine & Pharmacotherapy (2022)
Nunes, A. A., Dos Santos, R. G., Osório, F. L. et al. · Journal of Psychoactive Drugs (2016)
Ayahuasca is a complex psychoactive brew containing DMT and β-carboline alkaloids that can produce marked changes in perception, emotion, self-related processing, and large-scale brain connectivity. The introduction argues that although previous studies have described plasma alkaloids, metabolite changes, subjective effects, and functional connectivity in isolation, the multiscale biological relationships linking peripheral metabolism, brain network reorganisation, neurochemistry, and experience in humans remain poorly understood. The paper also notes that ayahuasca’s effects are not likely to be explained solely by 5-HT2A receptor activity, because downstream monoaminergic, lipid, and metabolic pathways may also be involved. Madrid-Gambin and colleagues set out to integrate multiple data layers collected before and after ayahuasca ingestion in experienced Santo Daime users. Their aim was to map how changes in plasma alkaloids and metabolomics relate to resting-state functional connectivity and to dimensions of the acute psychedelic experience measured with the 5D-ASC scale. By combining biochemical, neuroimaging, and psychometric data in a within-subject design, the study seeks to provide a systems-level account of the psychedelic state.
This was a within-subject, fixed-order observational study in 24 healthy adult volunteers who were long-term members of the Dutch Santo Daime community. Twenty participants with complete imaging and metabolomics data were included in the main analyses. Assessments were conducted on two consecutive days: baseline measurements on Day 1 and post-ayahuasca measurements on Day 2. Participants drank a mean of 24 mL of ayahuasca, and sessions were scheduled individually, around 90 min apart, at the same time of day across visits to reduce overlap of acute effects and circadian variation. Exclusion criteria included ferromagnetic implants, pregnancy, and use of medications within 24 h before testing. Subjective effects were measured with the 5D-ASC questionnaire, a 94-item self-report scale covering oceanic boundlessness, visionary restructuralization, anxious ego dissolution, auditory alterations, and reduction in vigilance. The questionnaire was completed at the end of the second study day. Resting-state fMRI was acquired about 60 min after dosing on a 7 T Siemens scanner. The authors used an in-house preprocessing pipeline with motion correction, nuisance regression, temporal filtering, and aCompCor-based physiological noise removal. Participants with mean framewise displacement above 0.5 mm were excluded, and scrubbing was applied for high-motion volumes; three participants were excluded for excessive motion. Connectivity was derived from 200 cortical regions and then reduced to Yeo seven-network summaries, including visual, somatomotor, dorsal attention, ventral attention, limbic, frontoparietal, and default mode networks. Proton MRS was collected between 70 and 90 min after dosing from posterior cingulate cortex and visual cortex voxels to quantify neurochemical ratios relative to total creatine. Plasma alkaloids were measured at 60 min after drinking by LC-MS/MS, including DMT, harmine, harmaline, and tetrahydroharmine. Targeted metabolomics was performed on plasma sampled 150 min after ingestion, using six LC-MS/MS methods to quantify 157 biomarkers across endocannabinoid-related compounds, amino acids, steroids, energy metabolism, acylglycerols, ceramides, lysophosphatidylcholines, and choline metabolism. For statistics, the authors used Wilcoxon signed-rank tests for pre-post metabolite changes and generalised linear models to relate within-person change scores to subjective dimensions, with false discovery rate control. The main integrative analysis used DIABLO, a multiblock partial least squares framework, with training and testing splits, cross-validation to select latent components, permutation testing for significance, and FDR correction. They also used regularised canonical correlation analysis to examine covariation between plasma metabolites and posterior cingulate cortex MRS, and between MRS and connectivity in the posterior cingulate cortex/default mode network and visual cortex/visual network pairings.
Of the 24 enrolled participants, 20 completed the full protocol. Three were excluded because of excessive head motion during fMRI and one because of insufficient biological samples. The ayahuasca brew came from a single batch and contained 0.14 mg/mL DMT, 4.50 mg/mL harmine, 0.51 mg/mL harmaline, and 2.10 mg/mL tetrahydroharmine. At the group level, ayahuasca ingestion was associated with significant increases in several endocannabinoid-related N-acylethanolamines, including anandamide (AEA), oleoylethanolamide (OEA), palmitoleoylethanolamide (POEA), and dihomo-γ-linolenoylethanolamide (DGLEA), after FDR correction. Smaller pre-post effects were also reported for some amino-acid-related measures, such as 5HIAA/5HT and leucine/LNAA, and for certain cortisone-related steroids, including 20α-dihydrocortisone and cortisone. No acylglycerol or ceramide/hexosylceramide measure survived FDR correction in the simple pre-post comparison. When individual metabolic responses were related to subjective experience, DGLEA and LEA were positively associated with oceanic boundlessness, auditory alterations, and visionary restructuralization, while POEA, OEA, and AEA also showed positive associations with selected 5D-ASC dimensions. In the amino acid block, auditory alterations and visionary restructuralization were inversely associated with several branched-chain and large neutral amino acid measures, especially leucine, isoleucine, tryptophan, phenylalanine, LNAA, and phenylalanine/LNAA. Glutamate and creatine were positively associated with oceanic boundlessness, and creatine was also positively associated with visionary restructuralization. Energy metabolism variables, including fumarate, citrate/malate, isocitrate, lactate/pyruvate, and related ratios, were significantly associated with at least two of the main subjective dimensions. Several acylglycerol and lipid changes were also linked to subjective ratings, including positive associations for MAG/DAG 18:2 and MAG/DAG 16:0 and negative associations for MAG/DAG 18:0; ceramide and hexosylceramide changes were mostly inversely associated, particularly with visionary restructuralization. In the multiblock model, one latent component was selected as optimal. This component carried the strongest loadings for auditory alterations, visionary restructuralization, and oceanic boundlessness, along with DMT, harmine, tetrahydroharmine, and several monoacylglycerol/diacylglycerol species, including MAG 19:0, MAG 18:2, DAG 18:2, DAG 18:0 18:2, and MAG 20:4 (2-AG). It also included endocannabinoid-related N-acylethanolamines, serotonin, and ceramide-related markers. For connectivity, dorsal attention–dorsal attention, somatomotor–dorsal attention, dorsal attention–default mode, and somatomotor–limbic edges were among the highest-loading features. The first latent component explained 47.1% of variance in the 5D-ASC block, 82.9% in alkaloids, 28.9% in functional connectivity, and 9.0% in metabolomics. Permutation testing reduced 6340 possible cross-block combinations to 610 significant associations. Only oceanic boundlessness, visionary restructuralization, and auditory alterations were significantly represented in the integrated network. All quantified alkaloids were highly connected hubs bridging peripheral and central data. Among energy metabolites, citrate/malate, alpha-ketoglutarate, succinate/alpha-ketoglutarate, isovalerate, malonate, and mevalonate were positively related to some dorsal attention, somatomotor, and default mode network edges and inversely related to others, with a cluster opposite the subjective-experience nodes in the network display. In the ASC-centred core network, the overlapping features linking all four layers included all plasma alkaloids, 10 acylglycerols, one hexosylceramide-related feature, and several connectivity edges, mainly involving dorsal attention, ventral attention, and default mode network pairs. Endocannabinoids, amino acid metabolism, and energy metabolism showed coordinated associations with alkaloids and connectivity, but their links to the ASC dimensions were only nominal and did not survive multiple-comparison correction in the final core network. Steroids, lysophosphatidylcholines, and choline metabolism were not significant in this integrative network. The rCCA analyses showed structured covariation between peripheral and central measures. In the posterior cingulate cortex, clusters of plasma lipids, including mono- and diacylglycerols and ceramides, were inversely associated with creatine, aspartate, NAAG, and glutathione, whereas phosphocreatine showed positive associations with these lipids. The plasma 5HIAA/serotonin ratio was positively associated with myo-inositol, NAA, and choline-containing compounds. A second rCCA linked posterior cingulate cortex NAA, NAAG, and glutamate/glutamine to within-default mode network and frontoparietal-default mode network connectivity, with the opposite pattern for sensorimotor, visual, and dorsal attention network connectivity; myo-inositol showed an inverse pattern. In the visual cortex analyses, only weak and inconsistent associations were found, and the authors did not interpret these further.
The authors interpret the findings as evidence that the acute ayahuasca experience reflects a multiscale brain-body reorganisation rather than a single neurochemical event. They argue that oceanic boundlessness, visionary restructuralization, and auditory alterations were the main experiential dimensions linked to coordinated changes in circulating alkaloids, peripheral metabolism, posterior cingulate cortex neurochemistry, and functional connectivity. They highlight the default mode network and dorsal attention network as especially prominent in the network reconfiguration, and they note that the normal anti-correlation between these systems appeared to weaken during the psychedelic state, consistent with a more integrated and less segregated brain organisation. Relative to earlier research, the authors say the results fit with established serotonergic accounts of psychedelics, particularly 5-HT2A-related cortical effects, but extend them by implicating endocannabinoid-related lipids, glycerolipids, sphingolipids, energy metabolism, and amino acid-related measures. They suggest that the associations with serotonin and glutamate remain broadly compatible with canonical psychedelic models, whereas the prominence of MAGs, DAGs, ceramides, and 2-AG points towards downstream lipid-signalling pathways that may accompany or support network reconfiguration and altered consciousness. They also note that the rCCA findings, linking peripheral lipids and serotonin-related measures with posterior cingulate cortex neurochemistry and default mode network connectivity, reinforce the view that peripheral and central changes covary in a structured way. The authors are careful to state that the data do not establish causality. They emphasise that the observed relationships may reflect parallel responses to alkaloid exposure, reciprocal brain-body interactions, or adaptive metabolic changes accompanying neural plasticity, rather than direct metabolic drivers of brain change. They also mention that mechanisms beyond 5-HT2A, including peripheral serotonergic pathways such as 5-HT2B, could contribute to the observed signatures. Key limitations include the modest sample size, the highly selected group of experienced Santo Daime users, the challenges of multimodal data collection in ceremonial settings, the acute cross-sectional design, and the fact that alkaloid concentrations were measured at only one time point. They note that longer time-course and pharmacokinetic-pharmacodynamic designs would better capture the dynamics of ayahuasca exposure and its biological effects. The authors also caution that the visual cortex MRS findings were weak and nonsystematic. They suggest future work should add longitudinal sampling, more detailed psychometrics or phenomenology, and richer real-time neural measures to better map how molecular and network dynamics relate to subjective experience and potential therapeutic effects.
The authors conclude that acute ayahuasca effects involve coordinated changes across alkaloid exposure, peripheral metabolism, posterior cingulate cortex neurochemistry, large-scale connectivity, and conscious experience. They state that, beyond established serotonergic models, endocannabinoid-related and other lipid pathways, especially acylglycerols and ceramides, appear to be prominent correlates of psychedelic phenomenology and network reconfiguration. They present the study as support for broader systems neurobiology approaches to altered states of consciousness.
A total of 24 healthy adult volunteers participated in a withinsubject, fixed-order, observational study. Data of 20 participants (11 males, 9 females; mean age = 37.1 ± 10.6 years) with complete imaging and metabolomics datasets entered data analysis. Participants were long-term members of the Dutch Santo Daime community, ensuring prior experience with ayahuasca. Assessments were conducted on two consecutive days. On Day 1, assessments were obtained at baseline (nonintoxicated state). On Day 2, assessments were obtained following ayahuasca ingestion (mean = 24 mL, SD = 8.16). Inter-individual variability in systemic exposure was addressed through pharmacokinetic assessment, by quantifying plasma concentrations of DMT and β-carbolines. Sessions were conducted individually and spaced approximately 90 min apart to minimize overlap of acute effects across participants. Testing was organized in four laboratory cycles of six participants each, with each participant assessed at the same time on both days to control for circadian variation. Exclusion criteria included the presence of ferromagnetic implants, pregnancy, and the use of medications within 24 h prior to testing. A full inclusion criterion and testing procedures are detailed in the Supplementary Materials. This study was conducted in accordance with the Declaration of Helsinki and its 2013 revision in Fortaleza, Brazil, and adhered to the Medical Research Involving Human Subjects Act (WMO). Ethical approval was obtained from the Medical Ethics Committee of the Academic Hospital and University of Maastricht (NL70901.068.19 / METC19.050). All participants provided informed consent after being fully briefed on study procedures, potential risks, and their right to withdraw without penalty.
Subjective effects were assessed using the 5-Dimensional Altered States of Consciousness Questionnaire (5D-ASC). This 94-item self-report instrument captures five core domains of altered states: oceanic boundlessness (OB), visionary restructuralization (VUS, from the original German "Visionäre Umstrukturierung"), anxious ego dissolution (AED), auditory alterations (AUD), and reduction in vigilance (VIR). Participants completed the questionnaire at the end of the second day by rating each item on a 10-cm visual analogue scale (VAS) anchored from 0% ("No, not more than usual") to 100% ("Yes, much more than usual").
The present study reuses data previously reported, but applies a novel data processing and integrative framework. Participants completed one imaging session per visit to collect resting-state fMRI data, approximately 60 min after dosing. To minimise circadian effects, all imaging sessions were conducted at the same time of day for each participant across visits. Images were acquired on a Siemens MAGNE-TOM 7 T MRI scanner. For complete details on image acquisition procedures and preprocessing, see Supplementary Materials. BOLD timeseries were preprocessed using an in-house pipeline based on established procedures, including motion correction, voxel-wise nuisance regression, and temporal filtering, incorporating aCompCor-based removal of physiological noise. Head motion was assessed using framewise displacement (FD), the temporal derivative of BOLD signal variance (DVARS), and signal standard deviation metrics, with participants exhibiting mean FD > 0.5 mm excluded from further analyses. Volumes affected by excessive head motion were identified and censored ("scrubbed") when framewise displacement (FD) exceeded 0.5 mm or when DVARS standardized by its temporal standard deviation exceeded the threshold defined as the 75th percentile plus 1.5 interquartile ranges. Three participants were excluded due to excessive motion. In the remaining sample, no significant differences in head motion were observed (see Supplementary Figure). Denoised BOLD time series were parcellated into 200 cortical ROIs using the 2 mm Schaefer atlasand mapped to native EPI space. Static functional connectivity was computed as pairwise Pearson correlations between regions and Fisher z-transformed. To reduce dimensionality, connectivity values were averaged within Yeo et al.'s seven canonical resting-state networks, including visual (VIS), somatomotor (SM), dorsal attention (DA), ventral attention (VA), limbic (L), frontoparietal (FP), and default mode network (DMN), retaining only unique network-level features for subsequent analyses.
Proton magnetic resonance spectroscopy ( 1 H-MRS) data were acquired in all participants to quantify in vivo brain metabolite concentrations between 70-and 90-minutes post-dosing. Single-voxel spectra were obtained from the posterior cingulate cortex and visual cortex using a stimulated echo acquisition mode (STEAM) sequence (TE = 6.0 ms, TR = 5.0 s, 64 averages)on a Siemens MAGNETOM 7 T MRI scanner. Spectroscopic voxels were positioned by a trained operator. MRS outcome measures consisted of concentration ratios of aspartate, N-acetylaspartate (NAA), glutamate, N-acetylaspartylglutamate (NAAG), GABA, glutamine, taurine, glycerophosphocholine, phosphocholine, glutathione, glycine, and myo-inositol relative to total creatine (creatine + phosphocreatine). The use of total creatine as an internal reference minimizes variability arising from transmit/receive RF inhomogeneity, magnetic field drift, and cerebrospinal fluid (CSF) partial volume effects within the voxel.
The present study is based on data previously reported. A blood sample was taken at 60 min post drinking to determine alkaloid concentrations in plasma. Plasma aliquots were frozen at -80 • C until analysis. Alkaloid concentrations were determined in 200 µL plasma using liquid chromatography-tandem mass spectrometry (LC-MS/MS) (Agilent, Waldbronn, Germany) after ethyl acetate extraction. Calibration curves covered the range 0.25 -40 ng/mL with lower limits of quantitation (LLOQ) of DMT 0.077, harmine 0.13, harmaline 0.23, and tetrahydroharmine 0,18 ng/mL.
Plasma samples were taken 30 min after the MRI scanning session (i. e. 150 min after ayahuasca ingestion) on both the baseline and ayahuasca study days. Following protein precipitation, samples were analyzed using six targeted LC-MS/MS methods covering distinct metabolite classes, as previously described. A set of 157 targeted biomarkers, composed of 12 compounds related to endocannabinoids, 25 markers related to amino acid metabolism, 13 steroids, 25 markers related to energy metabolism, 26 acylglycerols, 25 ceramides, 18 lysophosphatidylcholines (LPCs) and, 6 compounds of choline metabolism; were determined by selected reaction monitoring by LC-MS/MS system consisting of an Acquity UPLC instrument (Waters Associates, Milford, MA, USA) coupled to a triple quadrupole (TQS Micro, Waters) mass spectrometer. MassLynx software V4.1 (Waters Associates) was used for peak integration and data management. Large neutral amino acids (LNAA) were calculated as the sum of valine, leucine, isoleucine, tyrosine, tryptophan, and phenylalanine. Extended details are published elsewhere.
Demographic data were summarized as mean ± SD, counts, or percentages. All analyses were conducted in R version 4.4. Features with more than 80% missing values were excluded. Data distributions were evaluated both across data blocks and within individual features, and normality was assessed using Shapiro-Wilk tests. To assess the effects of ayahuasca on metabolite levels, fold changes were calculated and paired comparisons between baseline and post-ingestion values were performed using Wilcoxon signed-rank tests. For downstream integrative analyses, change scores (Δ) were computed for alkaloids, metabolomic variables, functional connectivity measures, and subjective experience by subtracting baseline values from post-administration values within each dataset. Additionally, generalized linear models with a Poisson distribution and log link function were applied to examine associations between metabolite changes (Δ) and subjective experience. Predictor variables consisted of Δ values (post-ingestion minus baseline), and models were fitted separately for each of the surviving 5D-ASC dimensions. Multiple comparisons were controlled using the Benjamini-Hochberg false discovery rate (FDR) procedure. The main data integration was conducted using the DIABLO algorithm (Data Integration Analysis for Biomarker discovery using Latent cOmponents), implemented in the mixOmics R package. A multiblock partial least squares (PLS) model was implemented, with the 5D-ASC dataset defined as the response block (Y-block). For model training and validation, autoscaled datasets were partitioned into training (75%) and testing (25%) subsets, with the 5D-ASC dataset specified as the response block (Y-block). The inter-block correlation in the design matrix was set to 0.5 to balance between discrimination and integration. Singular Value Decomposition (SVD) was independently applied to each block. The optimal number of latent components of each layer was selected by minimizing the root-mean-square error (RMSE) of predictions on the test set, using a modified implementation of the "predict.block.pls" function. Feature selection was conducted using a bootstrapping procedure, in which datasets were randomly shuffled and the multiblock-PLS model recalculated at each iteration. Statistical significance was determined by comparing the resulting similarity scores to a null distribution generated from 5000 permutations. The Benjamini-Hochberg procedure was used to control the false discovery rate (FDR), and features with an FDRadjusted p-value < 0.05 were considered statistically significant. A network graph was generated to visualize significant features and their interrelationships across data layers. Additionally, an ASC-centred core network was generated, including only nodes connected to at least one ASC variable and simultaneously linked to nodes in each of the remaining layers (quadripartite multi-layer network). Fig.illustrates the overall study design and data integration framework. In addition to the main modelling, regularized canonical correlation analysis (rCCA) was applied to explore associations involving an additional data layer, namely brain neurochemistry measured in two regions, in relation to peripheral metabolism and functional connectivity, using the mixOmics R package. Three separate rCCA models were implemented using delta-data (baseline-subtracted values), including (i) plasma metabolomics and posterior cingulate cortex MRS metabolites, (ii) posterior cingulate cortex MRS metabolites and DMN-related functional connectivity measures, and (iii) visual cortex MRS metabolites and visual network connectivity measures. Prior to analysis, all variables were mean-centered and scaled to unit variance. Regularization parameters were optimized using a grid search procedure with leave-oneout cross-validation, testing shrinkage values ranging from 0.001 to 1. Optimal parameters were selected based on cross-validated performance and subsequently used to fit the final rCCA models. For each analysis, pairwise correlation matrices were derived to quantify the strength of association between variables from the paired data blocks. These matrices reflect patterns of cross-block covariation captured by the canonical components and were interpreted as measures of statistical association rather than causal relationships.
Of the 24 participants enrolled, 20 completed the full study protocol, including blood sampling, resting-state fMRI, plasma metabolomic analysis, and subjective ratings of the psychedelic experience assessed with the 5D-ASC questionnaire. Three participants were excluded because of excessive head motion during fMRI, and one was excluded due to insufficient biological samples. Demographic and lifestyle characteristics of the final sample are summarized in Supplementary Table. Participants self-administered a single volume of ayahuasca equivalent to their usual dose (mean: 24 mL, SD: 8.16), prepared from a single batch by the Church of Santo Daime. The brew contained 0.14 mg/mL DMT, 4.50 mg/mL harmine, 0.51 mg/mL harmaline, and 2.10 mg/mL tetrahydroharmine. At the metabolic level, ayahuasca ingestion was associated with a marked modulation of the endocannabinoid-related Nacylethanolamine panel, with significant increases in anandamide (AEA), oleoylethanolamide (OEA), palmitoleoylethanolamide (POEA), and dihomo-γ-linolenoylethanolamide (DGLEA) after FDR correction (Supplemental Table). More limited pre-post effects were also detected in other metabolic blocks, including amino acid-related indices such as 5HIAA/5HT and leucine/LNAA (Supplemental Table), as well as cortisone-related steroids, particularly 20α-dihydrocortisone and cortisone (Supplemental Table). By contrast, no plasma acylglycerol or ceramide/hexosylceramide measure survived FDR correction in the pre-post comparison, indicating that ayahuasca did not induce robust group-level changes in these lipid classes (Supplemental Tableand).
We next examined whether interindividual metabolic responses to ayahuasca, using delta-data, were associated with the three ASC dimensions retained in the integrative model. In the endocannabinoidrelated N-acylethanolamine block, DGLEA and LEA were positively associated with OB, AUD, and VUS, while POEA, OEA, and AEA showed additional positive associations with selected ASC dimensions (Supplementary Table). In the amino acid block, AUD and VUS were inversely associated with several branched-chain and large neutral amino acid measures, particularly leucine, isoleucine, tryptophan, phenylalanine, LNAA, and phenylalanine/LNAA, whereas glutamate and creatine were positively associated with OB, and creatine was also positively associated with VUS (Supplementary Table). In the energy metabolism block, several metabolites and metabolite ratios were significantly associated with at least two of the three main ASC dimensions, including fumarate, fumarate/succinate, isocitrate, isocitrate/citrate, citrate/malate, lactate/pyruvate, and glutarate (Supplementary Table). Notably, despite the absence of robust grouplevel changes in acylglycerols and ceramide-related lipids after ayahuasca ingestion, several lipid Δ-values were significantly associated with ASC scores. Positive associations of MAG/DAG 18:2 and MAG/DAG 16:0, and negative associations of MAG/DAG 18:0, were observed across the ASC dimensions (Supplementary Table). Ceramide-and hexosylceramide-related changes showed predominantly inverse associations, especially with VUS (Supplementary Table).
For training the model, the optimal number of components in the multiblock-PLS analysis was determined by RMSE of the 5D-ASC dataset predictions in the test set. This optimization process identified one latent component as optimal for modelling with the lowest RMSE in the validation set (Supplementary Figure). In the validation set, this component showed that AUD, VUS and OB from the 5D-ASC block had the highest loading values before the inflection point ('elbow'), along with tetrahydroharmine, DMT and harmaline from the alkaloids block, and several monoacylglycerol (MAG) and diacylglycerol (DAG) species, such as MAG 19:0, MAG 18:2 (2-linoleoyl glycerol), DAG 18:2, DAG 18:0 18:2 and MAG 20:4, also known as the endocannabinoid 2-arachidonoylglycerol (2-AG), in the metabolomics block, as shown in Supplementary Figure. In this component, other endocannabinoid-related N-acylethanolamines appeared such as dihomo-γ-linolenoylethanolamide (DGLEA), linoleoylethanolamide (LEA) and oleoylethanolamide (OEA); serotonin (5-HT) and ceramides-related markers such as HexCer/Cer 20:0. Regarding the connectome edges, DA-DA, SM-DA, DA-DMN and SM-L showed the highest loading values, but a sharp decline of loadings values was not observed in this block. The retained first latent component of the final model explained 47.1% of the variance in the 5D-ASC block, providing additional context for the contribution of the multivariate model to individual differences in the acute subjective effects of ayahuasca. The same component accounted for 82.9% of the variance in alkaloids, 28.9% in functional connectivity, and 9.0% in metabolomics, supporting the interpretation that the principal multiblock axis was driven mainly by shared variation between exposure-related alkaloids, connectomic changes, and altered states of consciousness, with a smaller contribution from plasma metabolomic variation.
Permutation test-based feature selection denoted that from 6340 Fig.. Study design and multi-block data integration framework using latent components. The left panel shows a schematic of the study design, in which blood samples and fMRI data were collected from participants before and after ayahuasca consumption. The middle panel outlines the integrative analytical pipeline incorporating four data layers: brain connectivity (connectome), plasma alkaloids, global metabolomic signatures, and psychometric assessments of altered states of consciousness (ASC). Multi-block integration was performed using a latent component model within a cross-validated framework to capture shared variance across modalities. The right panel presents the resulting multiscale networks linking neurobiological, biochemical, and experiential dimensions across all blocks (upper) and at the centre of overlap (lower). combinations, only 610 remained significant after adjustment for multiple comparisons. Our multiblock integration model revealed that the four data layers (plasma alkaloids, metabolomics, functional connectome, and the psychedelic experience) were strongly interrelated (Fig.). The Circos plot (Fig.) illustrates the overall correlation structure across blocks, showing that only three 5D-ASC dimensions-OB, VUS, and AUD-were significantly represented in the integrated network. Among the biochemical variables, all quantified alkaloids (DMT, harmine, harmaline, and tetrahydroharmine) exhibited the highest connectivity degree across the datasets, occupying a central hub position that bridged the peripheral biochemical and central nervous system domains (Fig.). An interactive HTML version of the network, including all variable names and connection details, is available in the Supplementary Material. Among energy metabolism intermediates, citrate/malate, alphaketoglutarate, succinate/alpha-ketoglutarate, isovalerate, malonate, and mevalonate were positively correlated with DA-DA, SM-DA, and DMN-DMN connectivity, and inversely correlated with DA-DMN and Fig.. Multi-layer network of cross-block associations derived from the integrative multiblock-PLS model. (A) Circos plot visualizing all significant associations (FDR-corrected p < 0.05, based on permutation testing), with variables grouped by functional family. Metabolomic features are organized into major pathways, including endocannabinoids, ceramides, acylglycerols, amino acid metabolism, energy metabolism, and steroids, as detailed in Supplementary Table. Nodes represent individual variables, scaled by degree (number of cross-block connections), while edge colours denote the direction of association (red = positive, blue = negative), with line intensity reflecting the strength of correlation. (B) Integrative network plot displaying the overall topology and distribution of interconnected features. An interactive HTML version of the network, including all variable names and connection details, is available in the Supplementary Material. Abbreviations: 2-OG, 2-oleoylglycerol; AEA, anandamide; AUD, auditory alterations; VUS, visionary restructuralization; Cer, ceramide; DAG, diacylglycerol; DEA, docosatetraenoylethanolamide; DGLEA, dihomo-γ-linolenoylethanolamide; DHEA, docosahexaenoylethanolamide; DMT, N,N-dimethyltryptamine; DA, dorsal attention; EI, Elemental imagery; FP, frontoparietal; hexosylceramides; HexCer; MAG, monoacylglycerol; LEA, linoleoylethanolamide; L, limbic; LNAA, large neutral amino acids; LPC, lysophosphatidylcholine; OB, oceanic boundlessness; OEA, oleoylethanolamide; SM, somatomotor; VA, ventral attention; VIS, visual. VA-DMN links, forming a cluster topologically opposite to the ASC cluster in the network representation.
Next, we focused on the three ASC nodes as the central core connecting with the other three blocks, specifically, for each ASC node, we retained only those nodes in the other three blocks that showed a correlation with that ASC node and also exhibited a connection to nodes in each of the remaining blocks. Our findings highlighted a quadripartite link connecting the four data layers-alkaloids, metabolomics, functional connectivity, and ASC dimensions-through coherent cross-domain associations (Fig.). For OB and VUS (Fig.and), the network in which features from the four Fig.. ASC-centred core network. Quadripartite multi-layer network of cross-block associations derived from the integrative multiblock-PLS model. (A) Circos plot illustrating the quadripartite link across the four data layers: alkaloids, metabolomics, connectome, and ASC dimensions. Nodes represent individual variables, scaled by degree (number of cross-block connections), while edges indicate significant associations between layers. Panels B-D show integrative network plots highlighting the across-block associations with each of the ASC dimensions separately, revealing the topology of interconnected features. For clarity, metabolomic variables are grouped by compound family, as detailed in Supplementary Table. Edge colours denote the direction of association (red = positive, blue = negative), with line intensity reflecting the strength of correlation. Node size represents the number of significant connections per variable. Abbreviations: 2-OG, 2-oleoylglycerol; AEA, anandamide; AUD, auditory alterations; VUS, visionary restructuralization; Cer, ceramide; DAG, diacylglycerol; DEA, docosatetraenoylethanolamide; DGLEA, dihomo-γ-linolenoylethanolamide; DHEA, docosahexaenoylethanolamide; DMT, N,N-dimethyltryptamine; DA, dorsal attention; EI, Elemental imagery; FP, frontoparietal; hexosylceramides; HexCer; MAG, monoacylglycerol; LEA, linoleoylethanolamide; L, limbic; LNAA, large neutral amino acids; LPC, lysophosphatidylcholine; OB, oceanic boundlessness; OEA, oleoylethanolamide; SM, somatomotor; VA, ventral attention; VIS, visual. layers were intercorrelated simultaneously (overlapped features), included all plasma alkaloids, 10 acylglycerol and the formation of a HexCer lipid, and several connectome pairs mostly related either to DA, VA or DMN, including a prominent inverse association with DA-DMN. For AUD, the network included the same associations, and the formation of an additional HexCer (Fig.). Interestingly, although endocannabinoids, amino acid metabolism, and energy metabolism displayed coordinated associations with plasma alkaloids and functional connectivity features, their relationships with the ASC dimensions (OB, VUS, and AUD) were only nominally significant, but did not survive corrections for multiple comparisons in the integrative model and were therefore excluded from the quadripartite network (see Supplementary Tableand). No significance was found for steroids, lysophosphatidylcholines and choline metabolism.
rCCA integrating plasma metabolomics with posterior cingulate cortex MRS metabolites revealed structured patterns of covariation between peripheral and central measures. Distinct clusters of plasma metabolites, including lipid species (e.g., mono-and diacylglycerols, ceramides) and energy-related intermediates (lactate, malonate, mevalonate), showed differential associations with MRS-derived metabolites. MRS clusters 1 and 2, comprising creatine, aspartate, NAAG, and glutathione, were inversely associated with peripheral lipid species, whereas phosphocreatine (cluster 3) showed positive associations with these lipids. In addition, the plasma 5HIAA/serotonin ratio was positively associated with central myo-inositol, N-acetylaspartate, and choline-containing compounds (glycerophosphocholine + phosphocholine), shown in Fig.. In contrast, plasma neurotransmissionrelated compounds, including serotonin and amino acids, exhibited weaker and less specific associations with posterior cingulate cortex neurochemical measures. The corresponding regularized correlation matrix is provided in Supplementary Table(Fig.). Integration of posterior cingulate cortex MRS metabolites with DMNrelated functional connectivity measures revealed a structured organization of associations across connectivity patterns. Connectivity pairs involving within-network DMN and frontoparietal-DMN interactions exhibited similar association profiles across MRS variables, whereas those involving sensorimotor, visual, and dorsal attention networks showed an opposing pattern. MRS metabolites including NAA, NAAG, and glutamate/glutamine exhibited consistent associations with DMN-Fig.. Regularized canonical correlation analysis (rCCA) heatmap showing associations between plasma metabolite Δ-values and posterior cingulate cortex proton magnetic resonance spectroscopy (¹H-MRS) metabolite ratios. Rows represent plasma metabolites and columns represent posterior cingulate cortex MRS metabolites. Color intensity indicates the strength and direction of the regularized correlation coefficient. Hierarchical clustering was applied to rows and columns, and four column clusters are indicated above the heatmap. Abbreviations: 2-AG, 2-arachidonoylglycerol; 3OHKyn/Kyn, 3-hydroxykynurenine/kynurenine ratio; 5HIAA/5HT, 5-hydroxyindoleacetic acid/serotonin ratio; Asp, aspartate; Cer, ceramide; Cr, creatine; DAG, diacylglycerol; DEA, docosatetraenoylethanolamide; DHEA, dehydroepiandrosterone; GABA, gamma-aminobutyric acid; Glu, glutamate; Glyc, glycine; GPC, glycerophosphocholine; GPC+PCh, glycerophosphocholine plus phosphocholine; GSH, glutathione; HexCer/Cer, hexosylceramide/ceramide ratio; LPC, lysophosphatidylcholine; MAG, monoacylglycerol; NAA, N-acetylaspartate; NAA+NAAG, N-acetylaspartate plus N-acetylaspartylglutamate; NAAG, N-acetylaspartylglutamate; PCC, posterior cingulate cortex; PCr, phosphocreatine; POEA, palmitoleoylethanolamide. related connectivity, while myo-inositol showed an inverse pattern relative to these metabolites. The corresponding regularized correlation matrix is provided in Supplementary Table. In the visual cortex, rCCA revealed generally weak associations between MRS-derived metabolites and connectivity measures. Only two modest relationships were observed, namely between GABA and visual-attention connectivity and between taurine and visual-frontoparietal connectivity (data not shown), without consistent patterns across variables. Several metabolites were excluded because of insufficient spectral quality, and one participant was removed. Given the limited and nonsystematic nature of these findings, the visual cortex rCCA results were not interpreted further.
This integrative analysis provides a multiscale view of the ayahuasca experience, linking peripheral biochemical changes, posterior cingulate cortex neurochemistry, functional brain network reconfiguration, and alterations in subjective consciousness. By modelling within-subject Δ-values across alkaloid concentrations, plasma metabolomics, resting-state functional connectivity, and 5D-ASC scores, we identified biological shifts associated with key phenomenological dimensions of the psychedelic state. Specifically, OB, VUS, and AUD emerged as central experiential domains, each associated with distinct yet overlapping neurobiological signatures. The neurobiological fingerprints included all connectome networks, although edges involving the DMN and DA networks were overrepresented. In addition, they included metabolites related to energy metabolism, the endocannabinoid system, neurotransmission-related compounds such as glutamate and serotonin, and alkaloids present in the ayahuasca brew. Complementary rCCA analyses further supported a cautious interpretation of the peripheral findings by showing structured covariation between plasma metabolites, posterior cingulate cortex neurochemistry, and DMN-related functional connectivity. Several peripheral lipid species, including MAG-, DAG-, and ceramide-related measures, were inversely associated with posterior cingulate cortex markers such as creatine, aspartate, NAAG, and glutathione, whereas phosphocreatine showed the opposite pattern. In parallel, the plasma 5HIAA/serotonin ratio was positively associated with myo-inositol, NAA, and choline-containing compounds in the posterior cingulate cortex. A second rCCA indicated that posterior cingulate cortex metabolites including NAA, NAAG, and glutamate/ glutamine were positively related to within-DMN and frontoparietal-DMN connectivity, while showing the opposite pattern for dorsal attention-DMN, visual-DMN, and sensorimotor-DMN connectivity; myo-inositol showed an inverse profile. These findings do not imply that peripheral metabolites drive brain network changes, or vice versa, but they do suggest that ayahuasca-related peripheral metabolic responses are aligned with central neurochemical states that map onto the reconfiguration of core large-scale networks. The prominence of associations between the DMN and DA networks and the psychedelic experience under ayahuasca is notable. The DMN supports internally directed thought, including self-reflection, autobiographical memory, and maintaining a coherent sense of self. It is active when the mind is not focused on the external environment. In contrast, the DA network supports externally directed attention, helping to select and process sensory information relevant to current goals. These two networks typically show anti-correlated activity, meaning that when one is active, the other is suppressed. Their dynamic balance allows for flexible shifts between inner experience and engagement with the external world, forming a core organizational feature of normal waking consciousness. Under classic psychedelics, activity and functional connectivity within the DMN are reduced, particularly in hub regions such as the posterior cingulate cortex. This is much in line with the observation in the current study that changes within the DMN were associated with changes in the psychedelic state. At the same time, the usual anti-correlation between the DMN and the DA weakened during the psychedelic state with ayahuasca, suggesting that internal and external modes of processing became less segregated. This seems in line with the general notion that under psychedelics, global network connectivity increases, producing a more entropic and integrated pattern of brain activity. For example, although obtained with intravenous DMT rather than ayahuasca, recent multimodal EEG-fMRI work in humans showed convergent evidence of increased global functional connectivity, reduced network integrity, compression of the principal cortical gradient, and decreased alpha power, supporting the view that DMT-class psychedelic effects involve large-scale 5-HT 2A -linked reorganization of brain dynamics. Subjectively, this may correspond to intensified sensory experience, novel associations, and altered self-perception that are characteristic of psychedelic states. The integrative multiblock model revealed that these network reorganizations are biochemically coupled to specific lipid and metabolic pathways. The component explaining the largest shared variance linked OB, VUS, and AUD with circulating ayahuasca alkaloids, endocannabinoid-related N-acylethanolamines, energy metabolism intermediates, and lipid-derived metabolites, together with distributed changes in large-scale functional connectivity. Although these associations do not imply causality, the observed relationships with neurotransmission-related compounds such as serotonin and glutamate remain broadly consistent with canonical models of psychedelic action in which 5-HT 2A receptor activation engages glutamatergic cortical signalling. Associations with energy metabolism and amino acid-related neurotransmitters may reflect coordinated physiological processes rather than direct mechanistic effects. While previous work suggests that psychedelics can influence mitochondrial and oxidative activity in ways that support synaptic reorganization, the relationships observed here may alternatively reflect adaptive metabolic responses accompanying neural plasticity, rather than drivers of it. In addition, these associations could arise from peripheral or systemic processes, potentially mediated by serotonergic mechanisms beyond 5-HT 2A, including interactions with receptors such as 5-HT 2B , which are expressed in peripheral tissues and linked to metabolic and cellular signalling pathways. Thus, the observed metabolic signatures may reflect a combination of central neuroplastic processes and peripheral physiological responses, rather than a unidirectional effect of psychedelics on brain energy metabolism. Ceramide and acylglycerol lipids however have received little attention as relevant contributors to the psychedelic state before. Ceramides are bioactive lipids that modulate membrane microstructure, synaptic homeostasis, inflammatory tone, and are increasingly recognized as modulators of affective and cognitive states, while DAGs and MAGs represent key bioactive lipids in the endocannabinoid system. Interestingly, the MAG lipids also included 2-AG (MAG 20:4), reflecting the endocannabinoid lipid signalling during the psychedelic experience. DAGs are biochemical precursors involved in the synthesis of 2-AG. 2-AG signalling contributes to retrograde synaptic modulation, regulates excitatory-inhibitory neurotransmission, and shapes synaptic plasticity. The dynamic interplay between DAG and MAG (including 2-AG) provides an activity-dependent feedback system that adjusts synaptic strength and excitability of neural circuits. The negative relationships of psychedelic state features with DAG and MAG species suggest heightened activity of the 2-AG signalling cycle, where DAG is more rapidly converted to 2-AG by DAGL-α/β and MAG species reflect faster 2-AG breakdown. These findings are also in line with reports of elevated 2-AG mobilization and neuroplasticity following serotonergic stimulation. Conversely, positive relationships of HexCer/Cer 18:0 and 20:0 with psychedelic state and connectome features suggests that increments in ceramide metabolism are linked to synaptic network modulation and affective states. These findings extend serotonergic and glutamatergic models by showing that lipid-signalling pathways, particularly those involving glycerolipids and sphingolipids, covary with network flexibility and may influence large-scale functional brain organization. The coordinated variation between these lipid species, circulating alkaloids, and functional connectivity changes aligns with evidence that endocannabinoid-serotonin crosstalk supports neurophysiological plasticity. Complementary rCCA analyses further supported this interpretation by showing structured covariation between peripheral metabolites, posterior cingulate cortex neurochemistry, and DMN-related functional connectivity. In particular, several peripheral lipid species, including MAG-, DAG-, and ceramide-related measures, were inversely associated with posterior cingulate cortex neurochemical markers such as creatine, aspartate, NAAG, and glutathione, whereas phosphocreatine showed the opposite pattern. In parallel, the plasma 5HIAA/serotonin ratio was positively associated with myo-inositol, NAA, and choline-containing compounds in the posterior cingulate cortex. A second rCCA indicated that posterior cingulate cortex metabolites including NAA, NAAG, and glutamate/glutamine were positively related to within-DMN and frontoparietal-DMN connectivity, while showing the opposite pattern for dorsal attention-DMN, visual-DMN, and sensorimotor-DMN connectivity; myo-inositol showed an inverse profile. Although these associations are not causal, they suggest that peripheral lipid and monoaminerelated shifts under ayahuasca are aligned with central neurochemical states that map onto reconfiguration of core large-scale networks. When focusing on the overlapping network features linked to ASC dimensions, acylglycerols and ceramides emerged as the primary metabolic correlates, while neurotransmission-related compounds were no longer prominently represented. This pattern suggests a shift toward secondary biochemical processes downstream of receptor activation, with the shared neurobiological signature characterized more strongly by sphingolipid (ceramide) and glycerolipid (acylglycerol) signalling than by direct monoamine-related metabolites. The dominant connectome features remained stable and edges involving the DMN and DA network continued to represent central axes of network reorganization. The prominence of acylglycerols and ceramides in the shared ASCrelated network suggests that the subjective psychedelic experience may not only depend on primary neurotransmitter release but also on downstream lipid-mediated processes that reorganize large-scale functional connectivity of brain networks. In this framework, receptor modulation by DMT and harmines may initiate biochemical cascades in lipid signalling pathways that stabilize or amplify neural reconfiguration, thereby potentially contributing to altered states of consciousness. Overall, these findings support the notion that conscious experience is grounded in dynamic, system-wide integration rather than isolated neural events, incorporating peripheral processes in a manner consistent with contemporary neurophenomenological accounts of embodied consciousness. While this study provides an unprecedented integrative view of ayahuasca's multiscale effects, several limitations should be acknowledged. First, the sample size was modest, reflecting the challenges of conducting multimodal imaging and biochemical analyses in naturalistic ceremonial contexts. Access to this type of biological material and highly synchronized multimodal data remains logistically and ethically complex, which constrains replication and large-scale validation efforts. Moreover, as all participants were experienced users, the present findings may reflect adaptations associated with repeated ayahuasca exposure. Responses in novice or clinical populations could differ in both direction and magnitude. Second, while previous work suggests that psychedelics can influence mitochondrial and oxidative activity in ways that support synaptic reorganization, the relationships observed here may reflect adaptive metabolic responses accompanying neural plasticity, rather than its drivers. In addition, these associations could arise from peripheral or systemic processes, potentially mediated by serotonergic mechanisms beyond 5-HT 2A, including interactions with receptors such as 5-HT 2B , which are expressed in peripheral tissues and linked to metabolic and cellular signalling pathways. Thus, the observed metabolic signatures may reflect a combination of central neuroplastic processes and peripheral physiological responses, rather than a unidirectional effect of psychedelics on brain energy metabolism. Although the additional rCCA analyses revealed structured covariation between peripheral metabolites, posterior cingulate cortex neurochemistry, and DMN-related functional connectivity, (Fig.; Supplementary Tables), these analyses remain correlational and do not establish causal directionality. Accordingly, the observed associations may reflect parallel responses to alkaloid exposure across peripheral and central systems, reciprocal interactions between them, or a combination of both. Third, despite measuring plasma alkaloid levels at 60 min post drinking, the inherent complexity of the ayahuasca brew and the interinterindividual variability of the amount of ayahuasca consumed may introduce variability in drug concentrations and interactions with peripheral/central metabolites and brain imaging measures that might be better captured in pharmacokinetic-pharmacodynamic models that include multiple measures over the entire time window of the ayahuasca experience. Methodologically, the cross-sectional approach captures acute effects but does not address temporal dynamics or longer-term adaptations. Future work should incorporate longitudinal sampling to track recovery, persistence, or plasticity in both peripheral metabolomic and neural connectome signatures. Finally, while the current integrative framework bridges peripheral and central processes, the experiential dimension remains inherently complex. Advancing this line of research will require combining computational modelling, real-time neural recordings, and fine-grained psychometric or phenomenological assessments to better map how molecular and network-level dynamics give rise to subjective experience. Such multimodal, systems-level approaches will be critical for elucidating the neurobiological foundations of consciousness and informing the therapeutic potential of psychedelics. In conclusion, the present findings indicate that the acute effects of ayahuasca reflect a multiscale brain-body reorganization linking alkaloid exposure, peripheral metabolism, posterior cingulate cortex neurochemistry, large-scale functional connectivity, and conscious experience. Beyond established serotonergic models, our results identify endocannabinoid-and lipid-related pathways, particularly acylglycerols and ceramides, as prominent correlates of psychedelic phenomenology and network reconfiguration. The observed alignment between peripheral metabolites, posterior cingulate cortex neurochemical profiles, and DMN-related connectivity further supports the value of integrative multimodal approaches for characterizing how altered states of consciousness emerge from coordinated interactions across molecular, regional, and systems-level domains. This work therefore advances ayahuasca research from a primarily receptor-centred account toward a broader systems neurobiology of psychedelic experience.
Create a free account to open full-text PDFs.
Valle, M., Maqueda, A. E., Rabella, M. et al. · European Neuropsychopharmacology (2016)
Hesselgrave, N., Troppoli, T. A., Wulff, A. B. et al. · PNAS (2021)
Nicol, G. E. · Nature (2024)
Gattuso, J. J., Perkins, D., Ruffell, S. G. D. et al. · International Journal of Neuropsychopharmacology (2022)
Lebedev, A. V., Kaelen, M., L€ Ovd En, M. et al. · Human Brain Mapping (2016)
Mallaroni, P., Mason, N. L., Kloft, L. et al. · NeuroImage (2024)
Ramaekers, J. G., Mallaroni, P., Kloft, L. et al. · Journal of Cognitive Neuroscience (2023)
Studerus, E., Gamma, A., Vollenweider, F. X. · PLOS ONE (2010)
Tolle, H. M., Farah, J. C., Mallaroni, P. et al. · Network Neuroscience (2024)
Carhart-Harris, R. L., Friston, K. J. · Pharmacological Reviews (2019)
Timmermann, C., Roseman, L., Haridas, S. et al. · PNAS (2023)
Vollenweider, F. X., Preller, K. H. · Nature Reviews Neuroscience (2020)
Mason, N. L., Kuypers, K. P. C., Reckweg, J. et al. · Neuropsychopharmacology (2020)
Nichols, D. E. · Pharmacological Reviews (2016)
Martin, D. A., Nichols, C. D., Nichols, Á. C. D. · Current Topics in Behavioral Neurosciences (2017)
Milliere, R., Carhart-Harris, R. L., Roseman, L. et al. · Frontiers in Psychology (2018)