Major Depressive Disorder (MDD) [1] is a complex and widespread mental health condition, affecting millions globally. Characterized by persistent sadness, hopelessness, and anhedonia (lack of pleasure), MDD poses a significant challenge not only for the ones living with the condition but also for researchers and clinicians striving to treat it. Traditional approaches to treating MDD often fall short due to the disorder’s wide range of symptoms, causes, and treatment responses. However, recent advances in computational neuropsychiatry [2] are paving the way for more personalized and effective interventions.
Our team has explored these advances extensively in a recently published paper [3]. The key highlights include:
- Algorithmic Agent Model: A framework that views depression as a dysfunction in the brain’s Objective Function, leading to low emotional valence.
- Personalized Biotypes: Depression consists of distinct biological subtypes, each requiring tailored treatment.
- Functional Networks: Mapping specific brain circuits and networks linked to MDD symptoms.
- Neurotwins: Personalized brain models that simulate and optimize treatment strategies.
- Precision Psychiatry: Computational models enabling more targeted, individualized mental health care.
The Algorithmic Agent Model: A New Perspective on MDD
At the heart of this new approach is the concept of the algorithmic agent1, a theoretical model that views the brain as an information-processing system. This model is grounded in the Kolmogorov Theory (KT) of consciousness [4], which proposes that the brain operates as an “information compressor,” generating structured experiences2 by interacting with the external world. In this context, depression is seen as a state where the brain’s ability to generate positive valence (or emotional value) is persistently impaired.
1Algorithmic agent: Information-processing system that interacts with the world, creates models and acts to maximize its Objective Function.
2Structured Experience: Agent’s ability to interact with the world using accurate, encompassing, and compressed models. It occurs when the agent successfully compares model-generated data with real-world data, with the structure of the experience shaped by the characteristics of the model or program in use.
In the context of computational neuropsychiatry, depression can be viewed as a pathological state where an agent’s Objective Function—the system that evaluates valence—is persistently low. This model frames depression as a dysfunction in the agent’s ability to process and respond to the external world, reflecting an evolutionary design where the mechanisms that shape our cognitive systems are susceptible to maladaptive outcomes under certain conditions.
An agent, whether biological or artificial, can be broken down into several key components that interact dynamically to maintain emotional and functional stability. Understanding how these components malfunction helps us identify the potential causes of depression. Here’s a breakdown of the agent’s primary modules:
- The Modeling Engine builds and refines the agent’s internal world representation by predicting events and aligning actions with reality. Dysfunction here, like cognitive biases, can lead to misjudgments and emotional instability.
- The Simulator allows the agent to foresee scenarios. If malfunctioning, the agent may struggle to imagine positive outcomes, fostering hopelessness.
- The Comparator checks predictions against reality, refining the model. A faulty Comparator distorts perceptions, leading to emotional detachment or derealization.
- The Objective Function guides decision-making by evaluating scenarios to boost valence. When impaired, the agent remains trapped in a low-valence state, reflecting the stasis seen in depression.
- The Planning/Action Engine formulates actions based on inputs. In depression, it may fail to generate effective plans, leaving the agent feeling stuck.
- The World provides external signals, and a hostile environment can lower valence. Resilient agents recalibrate, while those with dysfunctional Objective Functions remain in a depressed state.
The Translational Framework: Linking Brain Circuits, Functional Networks, and Biotypes
The translational framework of the agent model consequently forms an interface between the high-level theoretical understanding and low-level practical application, by linking brain circuits, functional networks, and biotypes in MDD.
Brain circuits are networks of regions connected by structures like synapses, facilitating information exchange. Functional networks reflect the real-time activity of these circuits, showing how brain regions coordinate during tasks or rest. Biotypes, meanwhile, are distinct subtypes of depression, each defined by unique neurobiological traits that manifest differently across individuals, leading to variations in symptoms, treatment responses, and disease progression. These often emerge from patterns in functional networks (see [5]), which in turn reflect disruptions in underlying brain circuits. For example, a patient exhibiting cognitive dysfunction may have a biotype linked to abnormalities in the Cognitive Control Network, which is derived from circuits connecting the prefrontal cortex and other regions involved in executive function. By identifying which functional networks are altered, we can trace these abnormalities back to specific brain circuits. For example, disrupted connectivity in the Default Mode Network could indicate a biotype associated with excessive rumination, while deficits in the Mesolimbic Dopaminergic Pathway may point to anhedonia.
The agent framework is particularly helpful in this context, as it allows us to map the dysfunctions of these functional networks and circuits to core agent’s components. By targeting the specific circuits and networks disrupted in each biotype, we can tailor interventions—whether brain stimulation, psychotherapy, or pharmacological treatments—more effectively, paving the way for precision psychiatry.
The Future of MDD Treatment: Towards Computational Neurotherapeutics
The treatment of MDD requires a nuanced and personalized approach, given the disorder’s complexity and the variety of ways it manifests in different individuals. A deeper understanding of MDD’s underlying mechanisms is essential for developing effective treatments. This understanding can help tailor interventions to an individual’s specific symptom profile and the root causes of their depression. However, achieving this level of personalization requires a sophisticated theoretical framework, capable of guiding research into the etiology, patient subtypes, and treatment strategies for MDD.
One promising solution is the development of neurotwins [6]—personalized, mechanistic models of the brain designed to simulate the neural dynamics of an individual or a specific biotype of MDD. These models act as digital twins, reflecting the unique brain structures and functions of a patient. Such models provide a detailed simulation environment where the effects of various treatments, such as brain stimulation, pharmacology (including psychedelics), or psychotherapy, can be explored and optimized in advance.
Mechanistic models of this kind are crucial for personalized medicine because they allow us to go beyond general observations and actually simulate how different treatments might impact an individual’s brain circuits and functional networks. These models integrate brain imaging data, electrophysiology, and patient-specific factors to provide targeted, data-driven interventions. Recent advances in brain stimulation technologies, such as multichannel systems and focused ultrasound, enable clinicians to target specific brain networks. However, without a theoretical framework like the agent model to guide these interventions, it’s impossible to fully explore the vast array of treatment possibilities.
By combining computational modeling with emerging technologies, we are moving toward a future where treatments for MDD are not only more effective but also personalized to the neural and functional characteristics of each individual. This shift toward personalized neurotherapeutics represents a significant advancement in psychiatry, offering hope for better outcomes in managing this complex disorder.
References:
- [1] Bains N, Abdijadid S. Major Depressive Disorder. [Updated 2023 Apr 10]. In: StatPearls [Internet]. Treasure Island (FL): StatPearls Publishing; 2024 Jan-. Available from: https://www.ncbi.nlm.nih.gov/books/NBK559078/
- [2] Friston, K. Computational psychiatry: from synapses to sentience. Mol Psychiatry 28, 256–268 (2023). https://doi.org/10.1038/s41380-022-01743-z
- [3] Ruffini, G., Castaldo, F., Lopez-Sola, E., Sanchez-Todo, R., & Vohryzek, J. (2024, March 14). The algorithmic agent perspective and computational neuropsychiatry: from etiology to advanced therapy in major depressive disorder. https://doi.org/10.31234/osf.io/eqpjh
- [4] Ruffini G. An algorithmic information theory of consciousness. Neurosci Conscious. 2017 Oct 12;2017(1):nix019. doi: 10.1093/nc/nix019. PMID: 30042851; PMCID: PMC6007168.
- [5] Tozzi, L., Zhang, X., Pines, A. et al. Personalized brain circuit scores identify clinically distinct biotypes in depression and anxiety. Nat Med 30, 2076–2087 (2024). https://doi.org/10.1038/s41591-024-03057-9
- [6] Neuroelectrics Blog Post – Neurotwins (tES 2.0): Advancing Brain Simulation