This data-analytic study compared the similarity between several different psychedelic compounds, in terms of their reported subjective effects, binding affinity profiles, and molecular structures. Through the application of a novel machine-learning algorithm to the experience reports sampled from Erowid, the authors found that differences in subjective experience could be predicted by target binding site affinity and/or their conformational receptor states of the respective molecules. Notably, the 5-HT receptor subtypes yielded relatively poor predictions by themself in contrast to dopamine receptors (D1–5), which highlights that the dopaminergic action of LSD (in contrast to psilocybin) may elicit different types of subjective experiences.
Abstract
“Introduction: Classic psychedelics are substances of paramount cultural and neuroscientific importance. A distinctive feature of psychedelic drugs is the wide range of potential subjective effects they can elicit, known to be deeply influenced by the internal state of the user (“set”) and the surroundings (“setting”). The observation of cross-tolerance and a series of empirical studies in humans and animal models support agonism at the serotonin (5-HT)2A receptor as a common mechanism for the action of psychedelics. The diversity of subjective effects elicited by different compounds has been attributed to the variables of “set” and “setting,” to the binding affinities for other 5-HT receptor subtypes, and to the heterogeneity of transduction pathways initiated by conformational receptor states as they interact with different ligands (“functional selectivity”). Here we investigate the complementary (i.e., not mutually exclusive) possibility that such variety is also related to the binding affinity for a range of neurotransmitters and monoamine transporters including (but not limited to) 5-HT receptors.
Methods: Building on two independent binding affinity datasets (compared to “in silico” estimates) in combination with natural language processing tools applied to a large repository of reports of psychedelic experiences (Erowid’s Experience Vaults), …
Results: … we obtained preliminary evidence supporting that the similarity between the binding affinity profiles of psychoactive substituted phenethylamines and tryptamines is correlated with the semantic similarity of the associated reports. We also showed that the highest correlation was achieved by considering the combined binding affinity for the 5-HT, dopamine (DA), glutamate, muscarinic and opioid receptors and for the Ca+ channel. Applying dimensionality reduction techniques to the reports, we linked the compounds, receptors, transporters and the Ca+ channel to distinct fingerprints of the reported subjective effects. To the extent that the existing binding affinity data is based on a low number of displacement curves that requires further replication, our analysis produced preliminary evidence consistent with the involvement of different binding sites in the reported subjective effects elicited by psychedelics.
Discussion: Beyond the study of this particular class of drugs, we provide a methodological framework to explore the relationship between the binding affinity profiles and the reported subjective effects of other psychoactive compounds.”
Authors: Federico Zamberlan, Camila Sanz, Rocío Martínez Vivot, Carla Pallavicini, Earth Erowid & Enzo Tagliazucchi
Summary
Psychedelics are psychoactive substances that can elicit a wide range of idiosyncratic effects on consciousness, perception, emotion and cognition. They have been used ceremonially for millenia, and the discovery of LSD in 1943 signaled a period of intense scientific investigation.
Experimental evidence suggests that psychedelic effects are mediated by agonism at serotonin (5-HT)2A receptors, with a possible role for agonism at other 5-HT receptor subtypes such as 5-HT2C and 5-HT1A. This hypothesis agrees with the observation that cross-tolerance between classic psychedelics can be considered as evidence of a common mechanism.
The objective of this work is to investigate the variety of subjective effects elicited by different psychedelic molecules, in particular those elicited by relatively novel synthetic derivatives of phenethylamines and tryptamines.
The variety of subjective effects elicited by serotonergic psychedelics presents a challenge to the single-receptor hypothesis of psychedelic action. A correlation between the similarity of reported subjective experiences and the similarity of binding affinity profiles suggests that different receptors are involved in the reported subjective experiences.
Analysis Overview
Figure 1 shows the outline of our analysis, which includes data on the reported subjective effects of the compounds, data on the molecular structure of the compounds, and data on the binding affinity profiles of the compounds.
NIMH Psychoactive Drug Screening Program (PDSP)3 and the University of Basel in Switzerland obtained similar binding affinity profiles for drugs.
Binding Affinity Similarity
Data on the binding affinity of 18 psychedelic molecules were obtained from Ray (2010) and Rickli et al. (2015).
Two primary sources of data comprise the binding affinities published in Ray (2010). For LSD and ibogaine, data was obtained from the literature, while for the remaining molecules, new PDSP binding assays were conducted using the methodology described in Ray (2010) and Glennon et al. (2000). Ray (2010) described the procedure for normalizing K i values, and a logarithmic transformation with a sign inversion was applied. The lowest possible pK i value was assigned in circumstances where the primary assay did not produce >50% inhibition.
Rickli et al. (2015, 2016) performed radioligand binding assays at membrane preparations of human embryonic kidney cells and determined the affinity values of various ligands for 5-HT receptors, dopamine receptors, adrenergic receptors, the histamine H1 receptor, trace amine-associated receptors, SERT, DAT and NET.
To compare the binding affinities of two psychedelic compounds, we computed the Pearson’s linear correlation coefficient between the corresponding sets of pK i values.
Selection of Psychedelic Compounds
A subset of the molecules investigated in Ray (2010) was included in the present study. These partially overlap with those investigated in Rickli et al. (2015, 2016), and LSD was classified as a tryptamine for the purposes of this study.
Mescaline (3,4,5-Trimethoxyphenethylamine) is a naturally occurring substituted phenethylamine. It was first synthesized in 1919 by Spath, and 2C-B (4-Bromo-2,5-dimethoxyphenethylamine) was synthesized in 1974 by A. Shulgin.
A. Shulgin synthesized DOB, DOI, DOM, TMA-2 and DOM-2 in 1967, 1973, 1963 and 1962, respectively. These compounds act on the CB1 and CB2 receptors. MDA ([]-3,4-Methylenedioxyamphetamine) is a substituted entactogen phenethylamine, first synthesized in 1910 by G. Mannish and W. Jacobson and tested by G. Ailes in 1930. MDMA was not explored in humans until the 1970s. DMT, 5-MeO-DMT, 5-MeO-MiPT, DiPT, DPT, and LSD are naturally occurring substituted tryptamines, as well as endogenous to the human brain. They were first synthesized in 1931, 1936, 1985, 1979, 1981 and 1995, respectively.
Missing data for 5-HT2B, 5-HT1E, 5-HT5A, 5-HT6 , 5-HT7, D4 , D5, 1A, 1B, 2A, 2C, 2 , NET, I1 , M4 , 5 , H2 , CB1 , CB2 and Ca+.
A. Shulgin synthesized and tested several different types of psilocybin, including 2C-C, 2C-I, 2C-P, 2C-T-4, 2C-T-7, 25I-NBOMe, 4-OH-MET, 5-MeO-AMT, 4-OH-DiPT, and Psilocin.
We did not use reports of experiences with Psilocybe mushrooms as a proxy for the isolated active principle, because mushrooms can contain other psychoactive alkaloids, and because interoception may play a role in anticipatory events that can modulate behavior and subjective effects even before a drug has time to reach the brain.
Semantic Similarity of Subjective Effects Reports
The authors obtained reports of the subjective effects of different psychedelic compounds from Erowid’s Experience Vaults, which contains more than 20,000 reports. The reports were classified into different categories and the full corpus is described in a previous work. The data obtained from Erowid’s Experience Vaults was completely anonymous, and the measures followed to secure data confidentiality and privacy at the Erowid Center servers are described on its website.
The preprocessing of the reports was performed using the Natural Language Toolkit (NLTK) in Python 3.4.6. All words were converted to lowercase and lemmatized words containing less than three characters were discarded.
We applied LSA to quantify the semantic similarity between the reports of the psychedelic compounds included in this study. We used the term frequency-inverse document frequency (tf-idf) transform to compute a matrix in which rows are unique words in the corpus and columns represent ”documents”.
The word-by-document frequency matrix was decomposed into three matrices using Singular Value Decomposition (SDV). The first D largest singular values were retained and all others were set to zero, resulting in the reduced matrix of singular values S .
Structural Similarity
We computed atom pair similarities using the Tanimoto coefficient to measure the 2D structural similarity between the molecular structures of the compounds. The Tanimoto coefficient is a standard tool to quantify 2D molecular similarity.
Principal Components of Subjective Reports
We applied principal component analysis to the Erowid database and obtained five principal components that explained the highest variance in the tf-idf rank-reduced word frequency matrices. These five components were used to produce radar plots that quantified the presence of each receptor in the narratives.
Virtual Screening of Binding Affinity Profiles
We used a methodology introduced by Vidal and Mestres (2010) to obtain independent binding affinity values for a set of assayed active compounds.
Using the open-source data mining and integration platform KNIME, we extracted the molecular structures of compounds with their respective assayed binding affinity values from two public access chemical databases and calculated the similarity between the fingerprints of a pair of molecules using the Tanimoto distance coefficient.
A power parameter of 4.0 was used and a Layered algorithm was used to obtain a 39% of hits against 34 receptors, an acceptable result compared to the reference value of 59.3% – 66.7% achieved by Vidal and Mestres (2010) using the same procedure applied to a larger database of compounds.
Correlation Between the Similarity of Reported Subjective Effects, Binding Affinity Profiles and Molecular Structures
We first computed the similarities between the compounds based on the reported subjective effects, binding affinity profiles and molecular structures. We found that the Tanimoto coefficient differentiates phenethylamines from tryptamines, and that 2C-x and DOx compounds present high within-group structural similarity.
We observed positive correlations between binding affinity profile similarity and reported subjective effects similarity, between binding affinity profile similarity and molecular structure similarity, and between reported subjective effects similarity and molecular structure similarity.
The correlation between reported subjective effects and molecular structure/binding affinity profiles remained positive, but declined as a function of the number of components retained in the SVD step of LSA. A previously validated computational framework was also employed to estimate binding affinity profiles.
Replication of the Correlation Between the Similarity of Reported Subjective Effects and the Binding Affinity Profiles
We investigated the correlation between the semantic similarity of the reported subjective effects and the similarity of the binding affinity profiles of 2C-x compounds using the K i values published in Rickli et al. (2015, 2016). We observed high correlation between both variables except for the lowest number of dimensions.
Optimization of Receptor Selection for the Prediction of the Reported Subjective Effects and Molecular Structure Similarities
The correlations reported in Figure 4B correspond to those obtained using the binding affinities measured at all 42 sites. It is possible that higher correlations are achievable when considering only a subset of those sites.
We explored combinations of receptor types to find the one that maximized the correlations in Figure 4B. The optimal combination of sites was based on data from 5-HT, DA, opioid, muscarinic, NMDA receptors and the Ca+ channel.
We conducted a bootstrap analysis with 1,000 iterations, applying the LSA algorithm to 50% of the words in the Erowid reports. The results showed that 5-HT, DA, opioid, muscarinic, cannabinoid and NMDA receptors and the Ca+ channel were frequently included in the set that maximized the correlation.
Principal Components of the Reports of Subjective Effects
We applied PCA to the word-by-document rank-reduced matrix obtained after applying LSA to the tf-idf matrix and retained the first five principal components, which explained 58% of the variance of the data.
The variance explained by the 5 components is: body load, preparation, dependence, nausea, vomit, sick, money, puke, warm, tolerance, withdrawal, depression, anxiety, prescribe, vision, symptom, nausea, boil, medication, reality.
The principal components were named to indicate the nature of their most relevant terms. The dependence and therapeutic components included terms such as ”addict,” ”addiction,” ”withdrawal,” and ”tolerance,” as well as terms that could be related to the adverse reactions observed during the withdrawal syndrome of some addictive drugs.
Figure 9B shows the scores of the principal components for the 18 psychedelic compounds included from Ray (2010). The highest scores were observed for the perception component for 2C-E, DPT and DMT, for the body load component for 5-MeO-MIPT, TMA-2 and 2C-T-2, and for the dependence component for MDMA, ibogaine and 5-MeO-DMT.
Correlations Between Receptor Binding Affinities and Principal Component Scores
The correlation between the principal component scores of psychedelic compounds and their receptor binding affinities is presented in Figure 10A. The correlation between the principal component scores and the body load component is presented in Figure 10B.
DISCUSSION
We analyzed data on the reported subjective effects of 18 psychedelic compounds, their affinities at 42 possible binding sites and their molecular structure. We discussed the limitations of our study, with an emphasis on those related to the binding affinity data.
The semantic similarity matrix shown in Figure 4A and Figure 5 shows that compounds of similar molecular structure can produce similar subjective effects. There is scarcer evidence supporting such distinction for the other substituted tryptamines and phenethylamines analyzed in this study. Some of the most salient distinctions between classic psychedelics are related to their pharmacokinetics. Recent research suggests that the formation of particular ligand-receptor complexes influences the pharmacokinetics of LSD, as well as the activation of specific signaling pathways, which might impact on the potency and subjective effects of the drug.
While it is clear that 5-HT2A agonism underlies the psychedelic effect of certain substituted tryptamines and phenethylamines, the series of events that occur after the molecule interacts with the receptor is likely to depend on the conformation of the ligand-receptor complex. Activation of different signaling pathways by different psychedelic molecules could contribute towards establishing indirect links between the activated intracellular signaling events and the associated subjective effects. We found that the similarity of the reported subjective effects can be partially predicted by the binding affinity profiles at a wide range of receptors and transporter proteins, suggesting that functional selectivity at the 5-HT2A receptor may not be the sole contributor to the variety of reported subjective effects.
When considering the possibility that an ample range of receptors mediates psychedelic action, the analysis revealed that the action of LSD is mediated by 5-HT2A receptors, and that the action of psilocybin is mediated by D2 receptors.
Psychedelic tryptamines act on group II (mGluR2/3) and NMDA glutamate receptors, and may also affect 5-HT receptor-mediated behavior. They may also have effects on muscarinic and opioid receptors, although this area is largely unexplored.
The best prediction of molecular structure similarity between tryptamines and phenethylamines was achieved considering the binding affinities at 5-HT, imidazoline, muscarinic and receptors. This is consistent with the differential selectivity for 5-HT receptors between tryptamines and phenethylamines.
Published narratives lack numerical ratings of different items that characterize the psychedelic experience. PCA analysis identified components that could be identified with effects on consciousness, perception and bodily sensations, but not with substance dependence, therapeutic use and the preparation of psychoactive materials. The PCA analysis yielded results consistent with the known subjective and somatic effects of different drugs, and the link between the principal components and the binding sites reproduced some known facts about the functional role of different receptors.
The analogy with Fourier analysis is limited in the sense that the brain is a highly non-linear system and neuromodulation is not an exception. A typical example is the pharmacology of atypical antipsychotics, which perform well despite having lower D2 occupancy than first generation drugs.
It is uncertain whether the antagonized receptors can be extended to manipulate more subtle aspects of the effects elicited by psychedelic drugs.
Our study is based on the synthesis of data from diverse sources, including the Erowid database, and therefore inherits limitations intrinsic to each source of information. However, the large number of reports available from Erowid may allow a meaningful signal to emerge in spite of these uncontrolled sources of noise.
Particular emphasis must be given to the limitations inherent to the binding affinity data, which do not inform of the pharmacological action at the GPCRs. However, similar molecules may not present opposite actions at the same target, under the assumption that they occupy the same region in the binding sites.
We emphasize that binding affinity values may not be directly proportional to drug potency. The relationship is expected from a simple model of drug-receptor interaction based on the law of mass action.
Based on data included in Rickli et al. (2015, 2016) and Ray (2010), a significant correlation exists between the binding affinity similarities of 5-HT, DA and monoamine transporters and the reported subjective effects similarities. We note that dissociation constants reflect chemical/thermodynamical properties of the ligand-receptor interactions, whereas potency measurements are relative, e.g., to the choice of measured response. It is also not clear that in vitro measurements of EC50 /IC50 values are more accurate than binding affinities for the purpose of our analyses.
Given these limitations, our methodological developments yielded sensible results, and positive and significant correlations between reported subjective effects and binding affinity profile similarities were obtained from two independent sources of data. Future experiments should be carried out that contribute towards overcoming these limitations.
The scientific study of psychedelic drugs is a very young field of research, and human studies are unavoidable to fully understand their neuropsychopharmacology.
We have included all currently available sources of binding affinity data in our analysis, but we cannot avoid concluding that our study and its results should be considered as preliminary. Further replications of the binding affinity data measurements are required for a consensus to emerge, thus paving the way towards more robust and reliable results.
We proposed and applied a novel method towards the characterization of subjective reports of psychedelic compounds and their relationship to pharmacological data. This method is especially useful for the study of psychedelics, but could be used for other drugs as well.
AUTHOR CONTRIBUTIONS
FZ, CS, RMV, CP, FE, EE and ET analyzed data, prepared figures, and wrote the final version of the manuscript.
ACKNOWLEDGMENTS
We acknowledge Ariel Chernomoretz, Daniel Fraiman, Pablo Riera, Pablo Gonzalez, Facundo Alvarez, Juan Manuel Garrido, Luke Williams and David Nichols for comments and insights, and the founders, curators, contributors and volunteers of Erowid Center for sharing their data.
SUPPLEMENTARY MATERIAL
As discussed in a paragraph concerning the limitations of the pharmacological data, binding affinities do not necessarily predict EC50 and IC50 values for the 5-HT2A receptor and monoamine transporters.
Figure A1 shows that the correlation between binding affinity and drug potency is strong, and that the correlation between the binding affinities of the compounds assayed at 5-HT2A and 5-HT2C receptors is high (R = 0.90 for 5-HT2A and R = 0.86 for 5-HT2C).