For decades, psychiatry has relied on categorical silos—rigid diagnostic labels like "Schizophrenia," "Bipolar Disorder," or "Autism"—to classify mental illness. However, clinical reality is rarely so neat. Patients frequently present with overlapping symptoms, and high rates of comorbidity suggest that these traditional boundaries may not reflect the brain's underlying biological reality.
A growing body of research is shifting the focus from symptom-based labels to transdiagnostic brain signatures. This direction treats psychiatric and neurological disorders not as isolated entities, but as diverse manifestations of shared neurobiological disruptions. By identifying common structural and functional "fingerprints," researchers are moving toward a "biotype-informed" precision medicine that targets the mechanism rather than the label.
The Shared Structural Scaffold
The intuition behind transdiagnostic research is that the brain has a finite number of ways to fail. While the primary "trigger" might differ—a genetic aneuploidy in one case or a neurodegenerative process in another—the resulting damage often converges on the same critical hubs.
For instance, Franziska Knolle and colleagues demonstrated in Grey matter morphometric biomarkers for classifying early schizophrenia and PD psychosis: a multicentre study that psychotic symptoms in early schizophrenia and late-life Parkinson’s disease share a common structural signature in the thalamus. This suggests that "psychosis" is a conserved biological state that transcends the primary disease pathology.
Similarly, large-scale meta-analyses led by researchers like Pauline Favre have identified widespread white matter "fraying" in Bipolar Disorder that significantly overlaps with patterns seen in schizophrenia. These shared structural deficits, particularly in the corpus callosum and cingulum, act as a common biological scaffold for different clinical presentations.
Functional Convergence in Network Hubs
Beyond physical structure, the way brain regions communicate—functional connectivity—reveals even deeper commonalities. Research using "template-guided" Independent Component Analysis (ICA), such as the NeuroMark framework developed by Yuhui Du and Jing Sui, allows for the direct comparison of these networks across different studies.
In Evidence of shared and distinct functional and structural brain signatures in schizophrenia and autism spectrum disorder, Zening Fu and Yuhui Du found that Autism and Schizophrenia share significant abnormalities in sensory and cognitive networks. Interestingly, they observed a "spectrum of severity": ASD often presents as a quantitatively milder version of the same connectivity deficits found in Schizophrenia.
This convergence is particularly evident in the Default Mode Network (DMN) and the insula. As shown in a federated analysis by Kelly Rootes-Murdy in Cortical Similarities in Psychiatric and Mood Disorders Identified in Federated VBM Analysis via COINSTAC, gray matter loss in the bilateral insula is a "universal" marker across schizophrenia, depression, and autism. Because the insula processes emotional salience, its disruption may explain why diverse disorders all struggle with emotional regulation.
From Labels to Biotypes
If the underlying biology is shared, how do we explain the different symptoms? Researchers are increasingly using machine learning to move from categorical labels to dimensional biotypes.
The intuition here is that a single connectivity disruption can lead to different behavioral outcomes depending on a person's sex or developmental stage. Elvisha Dhamala and Sidhant Chopra demonstrated this in Brain-based predictions of psychiatric illness-linked behaviors across the sexes, showing that models trained to predict "acting out" (externalizing) in boys could predict "anxiety" (internalizing) in girls. The neural root is the same; the clinical expression is what differs.
To capture this, new frameworks like the Transdiagnostic Connectome Project (Sidhant Chopra et al., The Transdiagnostic Connectome Project: a richly phenotyped open dataset for advancing the study of brain-behavior relationships in psychiatry) and contrastive graph learning methods (Xuesong Wang et al., Delineating Transdiagnostic Subtypes in Neurodevelopmental Disorders via Contrastive Graph Machine Learning of Brain Connectivity Patterns) are being used to cluster patients based on brain connectivity rather than symptoms. These "data-driven" subtypes often contain individuals from multiple diagnostic groups (e.g., some ADHD and some ASD patients), but they share similar cognitive profiles and treatment responses.
Why This Works: The Pathological Gradient
The most striking evidence for this direction is the "lifespan spectrum" of brain disorders. Jingyang Zhang and colleagues showed in Deep learning reveals the common spectrum underlying multiple brain disorders that neural features from late-life Alzheimer’s can help detect early-life Autism.
This works because brain disorders likely exist on a continuous pathological gradient. Whether the cause is a rare genetic aneuploidy (Elizabeth Levitis, The variegation of human brain vulnerability to rare genetic disorders and convergence with behaviorally defined disorders) or urban environmental stress (Na Luo, Exploring risk and protective urban environmental factors on mental health through exposure network mapping), the brain responds by altering the same fundamental circuits of cognitive control and emotional processing. By mapping these transdiagnostic signatures, we move closer to a psychiatry that treats the brain's actual circuitry rather than its surface-level symptoms.