The Architectural Turn in Depression Why Network Topology Will Redefine CNS Strategy

 

By Denis Katz, MD, MHA
Founder, Salience Clinical

For decades, psychiatry has framed major depressive disorder (MDD) as a disorder of chemistry, an imbalance of serotonin, norepinephrine, or dopamine. That model produced meaningful therapeutic advances, yet it has not consistently delivered durable remission for a substantial proportion of patients. Roughly one-third remain treatment resistant, and among responders, relapse is common. Incremental modulation of synaptic signaling alone appears to have reached a ceiling in its ability to transform long-term outcomes at scale.

The next shift in neuropsychiatry will not be chemical.
It will be architectural.

By “architectural,” we mean the topology, coupling, and switching behavior of large-scale brain networks that shape how information is processed.

From neurotransmitters to networks

Advances in high-resolution functional MRI, multimodal imaging, and computational modeling over the past decade have transformed our understanding of depression. Major depressive disorder is now consistently associated with dysregulation across three large-scale brain systems, often referred to as the “triple network”: the Salience Network (SN), the Central Executive or Frontoparietal Network (CEN/FPN), and the Default Mode Network (DMN).

·        The Salience Network detects, integrates, and prioritizes internal and external stimuli, and supports dynamic switching between other large-scale networks.

·        The Central Executive/Frontoparietal Network governs high-level cognitive control, working memory, and goal-directed behavior.

·        The Default Mode Network supports self-referential processing, autobiographical memory, and, when dysregulated, maladaptive rumination.

What matters in this framework is not simply how strongly neurons fire, but how these networks interact and which network predominates at a given time. In MDD, studies report altered connectivity within and between DMN, CEN, and SN, including reduced anticorrelation between DMN and executive networks and abnormal salience–DMN/CEN coupling. In treatment-resistant subtypes, converging data point to more pronounced disturbances in these interactions, with altered connectivity patterns across salience-related, executive, and default mode circuits.

Converging evidence also points to impaired SN-mediated switching between the DMN and executive networks, contributing to persistent self-focus and insufficient recruitment of cognitive control. Together, these findings support a systems-level imbalance — an architectural problem in network coordination — rather than a purely local neurotransmitter deficit.

If MDD is increasingly understood as a disorder of triple-network coordination, the implications for how we design interventions are profound.

The strategic implication: signal modulation is not enough

Traditional pharmacotherapy primarily alters synaptic transmission. Neuromodulation platforms (such as rTMS, ECT, and invasive stimulation) primarily alter excitability in specific cortical or subcortical targets. But most existing approaches were not originally designed to deliberately recalibrate the balance and interaction of large-scale networks such as DMN, CEN/FPN, and SN.

We have been adjusting signal intensity. We have not been explicitly targeting network-level proportionality and switching behavior. If executive systems remain under-recruited and default/salience systems remain over-dominant or insufficiently inhibited, symptom suppression may occur without durable architectural recalibration, a plausible contributor to relapse vulnerability in some patients.

The emerging question for CNS innovation is therefore:
Can we design interventions that restore network reciprocity and flexibility, not just neurotransmission?

A shift toward network-level rebalancing

The future of precision psychiatry will likely include:

·        Imaging-informed patient stratification, using resting-state and task-based connectivity to define network-level biotypes and predict treatment response.

·        Circuit- and network-based endpoints in clinical trials, incorporating connectivity, network flexibility, and switching metrics alongside symptom scales.

·        Closed-loop neuromodulation guided by connectivity and biomarker signatures, as early work in treatment-resistant depression demonstrates the feasibility of biomarker-triggered stimulation in individual patients.

·        Plasticity-oriented biologics that enhance circuit resilience, with agents such as rapid-acting glutamatergic modulators appearing to reshape network connectivity alongside synaptic strength.

·        Behavioral and digital interventions explicitly designed to engage executive and regulatory networks while reducing maladaptive DMN dominance and aberrant salience tagging.

This is not a rejection of neurochemistry. It is a reframing of neurochemistry within systems neuroscience. Synapses operate within circuits. Circuits operate within networks. Networks define cognitive architecture.

Why this matters for pharma and medtech?

The architectural model reframes several strategic priorities for CNS innovation:

1.      Drug development
Compounds should be evaluated not only by their effects on symptom scales, but also by their impact on connectivity within and between DMN, CEN/FPN, and SN, and on network dynamics such as flexibility and switching. Network-level readouts can help differentiate mechanisms, refine dose selection, and identify responders. In practice, this means building connectivity and network-dynamics measures into early-phase studies rather than waiting until post hoc exploratory analyses.

2.     Clinical trial design
Network biomarkers and connectivity-based signatures may improve biotype stratification, enrich for likely responders, and reduce heterogeneity-driven signal dilution in both pharmacologic and device trials. Dynamic measures of network flexibility and switching may also serve as early markers of treatment trajectory. Programs that hard-wire these biomarkers into inclusion criteria, enrichment strategies, and key secondary endpoints will be better positioned for partner, payer, and regulatory conversations.

3.      Medical affairs narrative
The language of “chemical imbalance” is scientifically outdated and strategically limiting; contemporary data support a model centered on network- and circuit-level dysfunction, within which neurochemical modulation exerts its effects. A network-informed narrative better reflects current neuroscience and offers a more compelling rationale for integrated pharmacologic and device strategies.

4.     Long-term remission strategy
Durability may depend on restoring efficient switching, appropriate salience gating, and robust executive recruitment, rather than simply suppressing limbic or DMN hyperactivity in the short term. Strategies that explicitly aim to normalize or compensate for altered network topology may yield more sustained remission and reduced relapse.

A brief illustration

For example, a future treatment-resistant depression program might combine a rapid-acting modulator with rTMS targeted using pretreatment connectivity, and judge success not only by symptom change but by restoration of DMN–CEN–SN balance on imaging. That is what it looks like to operationalize architecture, rather than treat chemistry in isolation.

The competitive inflection point

We are entering a period where:

·        Network neuroscience has matured to the point that triple-network and broader connectomic models can inform development decisions.

·        Computational modeling and deep graph learning can extract predictive signatures of treatment response from multimodal brain networks.

·        Neuromodulation platforms are increasingly programmable and amenable to biomarker-guided, connectivity-informed targeting.

·        Multimodal datasets integrating imaging, EEG, and clinical phenotypes can support biologically grounded stratification and risk prediction.

The companies that integrate network architecture into their R&D, trial design, and lifecycle strategy are positioned to define the next decade of CNS therapeutics. Those that remain chemistry-only will increasingly compete in a domain of diminishing marginal returns.

The Salience Clinical perspective

At Salience Clinical, we operate at the intersection of systems neuroscience, clinical development strategy, biomarker integration, and translational positioning. Our focus is not theoretical neuroscience; it is applied architecture. We help teams translate emerging network science into concrete decisions about development pipelines, endpoint strategy, and evidence generation.

The next transformation in psychiatry will not come from louder drugs.
It will come from smarter maps.