Mapping the Mind with Psychedelics
Is there a role for a psychedelic 'data consortium'?
A recent study from researchers at Mcgill University, MIT, and SUNY Downstate came out last week. It is interesting for several reasons, but three stand out to me:
The ambitious goal of “Mapping the Mind” using psychedelics and other psychoactive drugs
The use of Artificial Intelligence and Machine Learning techniques to cross-reference thousands of trip reports and receptor-pharmacology data
A primary limitation—the reliability of the trip reports—strikes me as solvable as the emerging ‘infrastructure and operational’ layer of psychedelic medicine matures.
This study is the fourth1 effort that leverages the unique receptor-binding profile of psychedelic drugs—aka the “molecular fingerprint”— combined with first-person narrative accounts of the experiences—aka trip reports—to identify the specific pattern of brain activity that generates a particular subjective experience.
What I want to explore, and what is most germane for the ‘psychedelic ecosystem’ is how the explosion of interest in, and use of, psychedelics—in research, clinics, retreats, and personal use settings—could be leveraged to solve a significant limitation of this study and thus contribute to the goal of “Mapping the Mind.”
This methodology should only improve as more people use psychedelics and more clinics, researchers, and supportive technologies are developed to collect anonymized and encrypted experiential and semantic data with informed consent.
Let’s have a look.
What do we mean by Mapping the Mind?
From a summary published alongside the paper in the journal Science Advances:
“…to quantitatively marry real-world reports about psychedelic drug experiences, drug receptor binding affinities, gene transcription profiles, and the human brain…[this] method quantitatively translates a person’s experience into the molecular profiles responsible for that experience.
An excerpt from the abstract of Tom Ray’s 2010 paper Psychedelics and the Human Receptorome identifies the key idea:
“It should be possible to use this diverse set of drugs as probes into the roles played by the various receptor systems in the human mind.”
In other words, a ‘map of the mind’ means defining the precise combination of receptors, neural circuits, and brain regions that, when activated (or inhibited), are responsible for specific subjective states—thus knowing *how* to recreate these states on demand.
What is this research methodology?
The study design is succinctly captured in the press release issued by Mcgill (emphasis added):
“The researchers gathered 6,850 testimonials from people who took a range of 27 different psychedelic drugs. In a first-of-its-kind approach, they designed a machine learning strategy to extract commonly used words from the testimonials and link them with the neurotransmitter receptors that likely induced them. The interdisciplinary team could then associate the subjective experiences with brain regions where the receptor combinations are most commonly found…
the large testimonial dataset allowed the team to characterize coherent states of conscious experiences with receptors and brain regions across individuals…
"Our study provides a first step, a proof of principle that we may be able to build machine learning systems in the future that can accurately predict which neurotransmitter receptor combinations need to be stimulated to induce a specific state of conscious experience in a given person."
Why is this interesting?
The study is captivating because the intended result—a “map” of the brain regions responsible for specific subjective experiences—feels like it should require the most cutting-edge brain imaging technology or invasive Brain-Computer Interface (BCI) to monitor neural activity while subjects are actively under the influence of psychedelics in real-time.
But it doesn’t.
Instead, researchers used Natural Language Processing (NLP) techniques to process trip reports from Erowid Experience Vault of psychedelic experiences. Then, they cross-referenced this dataset with previously established receptor binding affinity data to generate a “Drug-Experience-Receptor Map.”
As the authors note:
“Each subjective drug experience was modeled as a specific combination of the brain-behavior factors. In this way, we have carefully deconstructed the changes of conscious awareness triggered by hallucinogenic drugs into its component parts…
we describe an anatomic grounding for the subjective experience of altered states of consciousness.”
In other words—and a dramatic oversimplification—this ‘map of the mind’ is created with machine learning/artificial intelligence techniques applied to two datasets:
The language people use to describe their experience of a specific psychoactive drug—i.e., trip reports (semantic dataset)
The specific neurotransmitter receptors to which drugs bind (pharmacological dataset)
It seems remarkable that such an ambitious product could come from such a seemingly simple methodology.
By simple, I mean that it is non-interventional, it is not a clinical trial, there is no bureaucratic red tape of procuring and handling scheduled substances, and most surprising (to me at least) does not use real-time, state-of-the-art brain imaging techniques.
And at first glance, this appears to be what the study offers.
However, despite the sophistication and novelty of the approach, I noticed criticism of both datasets.
First from Bryan Roth, who developed the psychedelic-receptor database used in these types of studies:
Then, Enzo Tagliazucchi, author of the second attempt at this approach, echoes Roth’s sentiment
And finally, Eduardo Schenberg highlights the limitation of the semantic dataset
So, there are two limitations, one from each dataset:
The substances (including the dosages, potency, and purity) people ingested cannot be verified, thus making the trip report data susceptible to error/inaccuracies. (In the final section of this article, I’ve outlined a few others)
The pharmacological data—Drug-Receptor Binding Affinities—is not specific enough. According to Roth and others, what is needed are agonist and antagonist2 potencies— that a drug binds to a specific receptor is only one factor—what activity results from this binding is what is most important.
Moving Forward: Pharmacology Data
In 2020, the Defense Advanced Research Projects Agency (DARPA) awarded Roth a $27 million grant, to create non-psychoactive drugs ‘inspired’ by psychedelics. But during an episode of the Mind and Matter Podcast, Roth further explains the details of the grant to establish a complete pharmacological profile of psychedelics:
“Roth: And so, in terms of the basic biology of psychedelics, we're actually looking at hundreds of known psychedelic drugs. And we're in the process of putting together basically paper that will serve as, as the database for our understanding of psychedelics. So every psychedelic drug that is known to be psychedelic, we're looking at right now. And at many different levels, so the receptor pharmacology, signaling, transcriptomics, proteomics.
Jikomes: So you're building a database, essentially, for researchers, where every known psychedelic will have a complete or at least a very thorough database of which receptors, they bind to what they do at those receptors, how genes are turned on or turned off because of that. And all of that will be available for researchers to tap into.
Roth: Yep. It'll be free online. For anyone.”
So, it seems the molecular work that Roth and colleagues are undertaking through the DARPA grant could address the limitation in the pharmacological data—but what about the limitations of the semantic dataset?
This is where the emerging ecosystem of clinics and ancillary technologies can contribute.
Moving Forward: Semantic Data
In this section, I want to think through how such semantic data may be collected going forward and how the ecosystem, now taking form could make this happen. Admittedly, I am pushing the limits of my understanding of data interoperability, governance and related matters—readers with expertise in this area are encouraged to respond with resources, guidance and suggestions.
It makes sense that the pharmacological data necessary for this approach to “mapping the mind” would come from a large government-sponsored research program because it is highly technical scientific work that requires the coordination of many trained scientists from different disciplines.
On the other hand, the semantic dataset came from thousands of people simply writing about their experiences with psychedelics after the fact.
These written trip reports from Erowid have been collected since 1995 according to the following criteria:
“Submissions should describe more than just a list of activities engaged in. We are looking for descriptions of mindset, dosage, physical and mental effects, intentions, insights, problems, drug interactions, after effects, etc.”
While many are well-written, carefully descriptive, and are thus valuable data points, there are a few limitations.
First, the purity and authenticity of the substances people consume are not verifiable. For example, people may assume they are taking MDMA or LSD, but unfortunately, contamination is common under prohibition.
Second, while many reports include dosage, route of administration, and setting—all of which significantly impact the subjective experience—such parameters may be better accounted for through structured and adaptive questionnaires.
Third, valuable data routinely collected in qualitative or clinical research settings, such as comorbidities, medications, medical history, diagnosis, etc., are not captured.
Fourth, these written accounts are produced after the experience and rely on participant’s recall. I sense that there would be value in collecting subjective experiences before, during, and after such powerfully mind-altering experiences—again, and this needs stressing—with robust privacy and anonymity measures and informed consent.
So, we’ve just noted four distinct areas that may improve the utility of the semantic data:
Validation of route of administration, setting, and other valuable parameters via explicit questionnaires
Inclusion of more formal health history
Semantic data from before, during, and after psychedelic experiences.
At the same time, there are two notable trends underway.
More people are using psychedelics3
An industry is forming to meet this demand
The industry taking shape includes policy reform efforts, retreat centers, psychedelic clinics, drug development projects, academic institutions dedicated to psychedelic research, and perhaps most importantly to this conversation, digital health applications.
As emerging health tech becomes more ubiquitous research, the practice of medicine, and consumer wellness are adapting.
For example, in the words of one FDA official, “Decentralized Clinical Trials are here to stay”
This emerging industry and the broad adoption of digital health technology could solve, to some degree, the limitations mentioned above.
Legalization, research, and clinical practice will necessitate the verification of substance purity, potency, and quality.
Smartphone-based research and measurement based care platforms could:
deliver structured and adaptive questionnaires about the substances, the setting, route of administration, etc.
allow subjects to upload or enter medical history, medications, and symptoms.
collect voice memos and journal entries before, during, and after psychedelic experiences.
Psychedelic Data Consortium?
While an update to the pharmacological data will require government funding and coordination of a major, multidisciplinary scientific effort, the semantic data could improve simply as a byproduct of the increasing use of psychedelics in conjunction with the technological infrastructure currently under development.
Getting this right will depend on verifiable and documented robust privacy, encryption and anonymity practices and informed consent.
To this end, I wonder if such an effort to coordinate data collection across the many stakeholders need a data consortium or commons of some sorts?
Perhaps a Psychedelic Research Data Consortium?
A Brief History of Trip/Mind Mapping
This paper is the latest attempt at Drug-Experience-Receptor Mapping.
My research indicates that the possibility of such a research design begins with the National Institute of Mental Health Psychoactive Drug Screening Program (NIMH-PDSP) completed by Bryan Roth, then of Case Western Reserve, currently at UNC.
“An agonist is a drug that binds to the receptor, producing a similar response to the intended chemical and receptor. Whereas an antagonist is a drug that binds to the receptor either on the primary site or on another site, which all together stops the receptor from producing a response.” (source)