In recent years, psychedelics have moved from spiritual ceremonies and music festivals to clinical trials for the treatment of addiction, PTSD and depression. Oregon and Washington DC have already taken steps to decriminalize certain psychedelics, and ketamine and psilocybin clinics have popped up in the United States.
“It’s a bit of the Wild West,” says Sam Freesun Friedman, a senior machine learning scientist at MIT and Harvard University’s Broad Institute.
But psychedelics are still largely illegal in the United States, in part because of their unpredictability. Reactions to different psychedelics vary widely: some users experience overwhelming healing or euphoria, while others come out with scarring trauma or terror. These and other reasons make it difficult for these drugs to be approved by government agencies and make it into doctors’ offices.
This week, Freesun and researchers from SUNY Downstate Health Sciences University and McGill University published an article in the journal Scientists progress to provide a unique method to better understand the interaction between hallucinogenic drugs, people’s brains and different types of psychedelic experiences. They did this by using artificial intelligence to examine real-life accounts of psychedelic experiences and compare them to how human brain chemistry interacts with drugs at the molecular level. However, although the researchers’ methods and goals push the boundaries of understanding how psychedelics can help or harm individuals, the data they use may not be reliable.
To collect descriptions of psychedelic trips from real people, the team used a non-profit website called Erowid which contains more than 40,000 anonymous user-submitted anecdotes about people taking psychoactive drugs. For the first set of data, the researchers extracted nearly 7,000 written Erowid accounts of 27 drugs, including LSD, ketamine, MDMA (also known as molly or ecstasy) and psilocybin (the compound active magic mushrooms). They then used a natural language processing tool to look for similarities in descriptive wording both within experiences with the same drug and between different drugs, Freesun says.
For the second set of data, the authors drew on previous research on how each psychedelic interacts with the human brain at the molecular level. Specifically, they looked at binding affinities, which quantify the ability of a drug molecule to bind to a particular neurotransmitter receptor. They then used a form of machine learning to find connections and patterns between the neurotransmitter receptors associated with each drug and the sensations people described while taking the substance.
Based on this analysis, Freesun and his collaborators found eight categories of receptor-experience combinations that he believes can be considered the Big Five personality traits for psychedelic experiences. Just as some might rate an individual’s personality on openness to experience, conscientiousness, extroversion, agreeableness, and neuroticism, researchers show how each drug or trip might rank on the spectrum of factors such as conceptual versus therapeutic, euphoria versus terror, and relaxation versus nausea.
[Related: What happens when psychedelics make you see God]
The implications of these findings envision a future where scientists could chemically modify a drug to achieve desired experiential effects for patients. For example, this approach could be useful for maintaining the therapeutic effects of a psychoactive drug while minimizing the terrifying experience typically associated with it, Freesun says.
“Finding a data-driven way to structure these experiences to maximize therapeutic benefits, I think is something we are all passionate about,” he adds.
But the foundation on which the study was built is flawed, says Bryan Roth, a professor of pharmacology at the University of North Carolina School of Medicine and director of the National Institute of Psychological Drug Testing Program. mental health (NIMH-PDSP). While Roth thinks the paper’s methods pose an “interesting insight,” he says the Erowid and biological data are unreliable, as well as the paper’s conclusions.
For starters, Roth says Erowid doesn’t check the chemical makeup of the drugs described in each story. “In a lot of cases, the drugs bought on the street are not the ones the person thought they bought, especially when it comes to psychedelic compounds and hallucinogens,” he explains. As an example, Roth points to how two US Military Academy cadets recently overdosed on cocaine that was actually mixed with fentanyl.
This, according to Roth, presents a problem when trying to draw connections between narrative data and how each drug behaves in the brain; a study could use the words of someone who took mislabeled MDMA and link it to the effects of real, pure MDMA. Erowid even has an independent laboratory that studies samples of drugs bought on the street. In 2021, it analyzed 747 drug samples sold as MDMA – a quarter of those samples contained other compounds or no MDMA at all.
Freesun agrees that illicit drugs may contain impurities or be mislabeled, but he says there is no reason to believe the inaccuracies are widespread enough to cast doubt on the article’s conclusions. His team checked the narrative data by stratifying it by sex and age to see if that skewed the results. They concluded that the subcategory results were still very consistent with the data set as a whole.
The second review hits a little closer to Roth. the Scientists progress the article quotes a 2010 PLOS One publication by Thomas Ray as one of two primary sources for his bond affinity correspondences. Ray’s paper relied on screening data from NIMH-PDSP, the lab run by Roth, but he says the information isn’t strong enough to analyze for further drug research.
“What we say [other scientists] that is, if they want to publish the data, we have to replicate it at least three times to make sure the values are correct,” says Roth. He notes that he told Ray that the NIMH-PDSP did not have the resources to replicate the data to prove its accuracy. Roth himself had chosen several incorrect values and therefore did not believe that binding affinities should be accepted as fact.
“He published it anyway,” he says. Freesun responds that his team was unaware of Roth and Ray’s conversation, but points out that more than 200 other articles cite the same data set.
But even if the binding affinity dataset was reliable, it’s not the right metric to use for the new study, Roth says. Binding affinities don’t show how well a drug activates a neurotransmitter receptor, he explains, so a compound could be classified as having low affinity for a receptor but still having very high potency. On the other hand, a psychedelic compound might have a strong affinity for a certain receptor but end up blocking it, Freesun says.
[Related: The tasty chemicals flavoring the edible cannabis boom]
Freesun also agrees that binding affinities don’t tell the whole story, and that using data with a more direct representation of how a psychedelic compound interacts with receptors would be a huge step forward for future research. However, he says the paper’s findings are still relevant and that the statistical and AI tools used by his team were deliberately chosen to filter out “noise” or inconsistencies in the data to find patterns.
“The study is driven by the question of what we can find despite [the noise]”, Freesun wrote in an email. “The large number of confirmatory results…has convinced us that there is a signal to be found amidst the noise.”