Social Physiology for Precision Psychiatry

Guillaume Dumas, MENG, MSC, PHD, HDR

FPR CMB Network Meeting

Moderated by Carol Worthman via Zoom

November 12, 2021

  • IVADO Assistant Professor of Computational Psychiatry, Department of Psychiatry and Addiction | University of Montreal
  • Director | Precision Psychiatry & Social Physiology Laboratory, CHU Sainte-Justine Research Centre
  • FRQS J1 Researcher in Artificial intelligence and Digital Health, Mila – Quebec Artificial Intelligence Institute
  • Website: extrospection.eu / ppsp.team Twitter: @introspection / @ppsp_team

Intro

  • Lab in three research areas: computational psychiatry, precision medicine, social neuroAI, including using AI as a bootstrap for hypotheses on social cognition.
  • Focus: Understanding psychiatric conditions across scales from genes to cells, circuits, organ systems, and organisms, and from different theoretical perspectives, e.g., Bayesian, cognitivist, dynamicist.
  • Also taking into consideration the different actors or stakeholders: patients, researchers, funders, clinicians, etc.

Social Interaction:  The “Dark Matter” of Neuroscience

  • Neuroscience until recently focused on individuals in isolation, e.g., single brain fMRI.
  • Over the past ten years, Dumas has been trying to understand and operationalize naturalizing social interactions between individuals, including how social interaction affects the brain and vice versa.
  • The social physiology of psychiatry combines social neuroscience and computational biology to offer a multiscale perspective. According to Berrios’s Cambridge model, psychiatric symptoms are the result of a “dialogical negotiation” illustrated as the outmost envelope of an inchoate biological signal enveloped by cultural configurators (Berrios, 2013).
  • A multiscale perspective addresses the “social cognition paradox”: “You need a mirror neuron system for social interaction but you need to interact to form a MNS” (Dumas, 2014).

SOCIAL NEUROSCIENCE

  • Hyperscanning: Inter-brain synchronization occurs through perception and action loops during social interaction: the two brains “couple” or appear to be acting as a single system (replicated in language interaction; Dumas et al., Cortex, 2019; and in affective touch)
  • In clinical settings, rather than a deficit, is autism a problem of dialectical misattunement between two (neurodiverse and neurotypical) brains?
  • Challenge is explaining the interbrain synchrony using experimentation and computational tools.
  • In a recent cable-less optogenetics study, Yang et al. zapped the mPFC of multiple mice at same time and same frequency to achieve interbrain synchrony (Yang et al., 2021, Nature Neuroscience). The results of that study suggested brain synchrony is a functionally meaningful and causal factor of social behavior.
  • Could there be a therapeutic application? (A current research question.)

COMPUTATIONAL BIOLOGY

  • The case of autism as a diagnostic category, now considered a spectrum.
  • Are subgroups evident from the data, or is autism heterogeneous?
  • Problem with grouping controls vs. patients: statistically significant does not equal clinically significant.
  • “Beyond the Gaussian curve”: distributions can be deeply misleading.
  • The challenge is to go beyond “cases vs. “controls” reductions, and look at continuities and network-based stratifications.
  • Another possibility, how do subgroups react to different treatments (CBT, drug, etc.)?
  • Normative models can be used to describe clinical trajectories: stratifying CTR, ASD, Intellectual Disability, ASD-ID to illustrate working memory score by age among these subgroups.
  • The point is, this work can disentangle “confounds that can lead to confusing and misleading conclusions.”
  • Needless to say, the language of mathematics is limited, but it is also open-ended, new mathematics can be invented.

One of the strengths of computational modeling is that it naturally draws out connections between levels. These connections can lead to insights in psychiatry. – David Redish & Joshua A. Gordon, Computational Psychiatry: New Perspectives on Mental Illness

THE THREE CULTURES OF COMPUTATIONAL PSYCHIATRY

  • Generative modeling (e.g., neuroeconomics, decision-making), can be agnostic re hypotheses.
  • Digital tools
  • Machine learning, a type of AI
  • Technologies: T1- and diffusion-weighted-MRI, f-MRI, EEG
  • Uses: Diagnosis, prognosis, treatment decisions
  • Models: statistical, biological informed, generative, social generative (Dumas, 2012, at dyadic level), and interactive (Dumas, 2014), to explore sensorimotor social coordination and representational social skills, i.e., the social cognition paradox.
  • Example re ASD: Dumas is using a human-machine interface (the Human Dynamic Clamp system) for interactive psychometric assessments in waiting room, prior to clinical visit (Dumas, 2019), and showed us a clip of a mixed reality session.

FINAL REMARKS

  • Interested in adding social dimensions to RDoC domains, but could be funding issues.
  • Network analysis paper in preparation (regarding autism and stakeholders), written during pandemic addresses the misalignment of topics of interest between researchers and patients/families, etc. (Gauld et al., 2021).
  • Aging is missing in autism research, and a major concern for families.

DISCUSSION

  • CMW: Deep phenotyping of social cognition question: Re role of context in driving development, Carol wondered if and how your work can build on studies of developmental processes? Dumas said it would be necessary to agree on a “taxonomy of context.” Using a network-based approach, you could incorporate life events that activate symptoms, which interact with other symptoms, and affect trajectories.
  • SBS: “Aren’t the concepts of ‘neurotypical’ vs. ‘neuro-divergent’ and ‘precision medicine’ diametrically opposed? The former involves typological thinking. So how can you get to precision medicine when the naturally occurring heterogeneity (neuro-diversity) is clouded by too few categories (i.e., typical vs divergent; autistic vs. normal).” Dumas suggests a stratification approach using subtypes, with possible differences among cultures. Patients are also almost never exactly what the DSM describes. A data-driven taxonomy and the DSM can perform complementary roles.
  • JS: Re positive and negative experiences in video-game play. There is a lot of hidden diversity at the level of culture group (i.e., those sharing patterns of learned thought and practices). Ethnographers can point, on the basis of a single case study, to specific culture-group effects, but where does that fit on a larger continuum of various kinds of experiences? Also, if you end up pointing to several sub-cultural systems, you’re not really explaining much. Going back to the slide “Beyond the Gaussian curve,” negative and positive effects of sub-groups could potentially cancel each other out.
  • GD: You are reducing to one dimension, something that is n-dimensional. The answer could be big data, would differences show up in networks?
  • LK: For me, the big concern is, we don’t know the dimensionality of social niches, i.e., we don’t know what the adequate number of variables is that we need to capture. Ethnography is hugely important here in terms of being able to identify a dimension that differentiates groups or people.
  • CMW: Referring to the problems with studying social ecology vs. the imaging and mathematical tools we have for studying the brain, perhaps it’s up to anthropology and the social sciences to bring the same kind of complex thinking and operationalization of ideas to get at these questions.
  • GD: We can bootstrap sociological or anthropological theories with social generative models and probe those questions in a quantitative way. Dumas’s lab is planning to do multi-agent simulations to understand effects at the group level. Regarding JS’s question of underlying mechanisms (referring to his pandemic lexical network analysis revealing differences in topics of interest to ASD researchers and to the general public based on data mining Twitter and PubMed; see Gauld et al., 2021), this would require major cooperation among disciplines, but could conceivably point to substantive real-world issues.
  • LK: Suggested a book co-authored by Sandro Galea, Growing Inequality: Bridging Complex Systems, Population Health, and Health Disparities. But, is the investment being made to construct the right parameters or dimensions?
  • CMW: Referring to the lexical network analysis, this is an interesting way to provoke research and to provoke cultural communication analysis to address the question, How do we speak to people’s real, felt needs?
  • GD: The analysis also revealed interesting topics among the general public, such as “canine companionship” or “money management,” with potential for “dramatic positive impact” on lives of people living with ASD.
  • LK: Another level, even though these are separate communities, they are interacting in the form of, e.g., advocacy groups. The definition of who has autism is a social as well as scientific process, with interactions between the two.
  • JS: Gave example of massive increase in BPD diagnosis in children between 1993-2004.
  • GD: In US, insurance coverage for an ASD diagnosis is greater than for an intellectual disability, totally biasing biological aspects.
  • Suggested Readings