Mental health disorders are a growing global crisis. According to a large-scale 2023 study spanning multiple countries (J. J. McGrath et al., Lancet Psychiatry 10, 668–681; 2023), one out of every two people in the world will develop a mental health disorder in their lifetime. Unfortunately, despite decades of research in psychology, neuroscience, and genetics, the underlying mechanisms of psychiatric disorders remain largely unknown, partly due to the complex interplay of genetic, environmental, and developmental factors, and the heterogeneity of symptoms across individuals. While advances in brain imaging, genomics, and behavioral assessment have generated large datasets that could be used to improve our understanding of the neurobiology and genetics of major mental health disorders, integrating these diverse sources of information into coherent models is a daunting task.

Credit: sorbetto / DigitalVision Vectors / Getty Images

Computational psychiatry, although still in its early stages, offers a promising avenue in the field of mental health, by leveraging computational models, machine learning, and data-driven insights to explore theories of cognition and to bridge the gap between neural mechanisms and clinical symptoms. While the models might not yet be ready for use in clinical settings, they could be employed by researchers and practitioners for gaining novel insights and supporting traditional psychiatry. It goes without saying, however, that these developments do not come without challenges that the research community must effectively understand and address.

On this World Mental Health Day, we present a Focus issue on computational psychiatry, where we highlight the convergence of psychiatry and computational science, as well as explore the many challenges that this integration brings.

As computational psychiatry seeks to move from theoretical modeling to clinical applications, several hurdles lie ahead. In a Comment, Quentin Huys and Michael Browning discuss the origins of the field and argue that it is moving towards causal approaches by using data and findings from clinical trials. The authors also highlight major barriers, such as outdated outcome measures, measurement reliability, and the divide between clinicians and researchers, and call for collaborative efforts to translate computational insights into clinical impact.

Since mental health research increasingly leverages sensitive personal data, protecting patient privacy has become another critical concern in the field. Iryna Gurevych and colleagues examine, in a Perspective, how researchers can build privacy-aware models of mental health, from using anonymization solutions to employing privacy-preserving methods, with the goal of creating secure, effective clinical support tools.

The growing integration of artificial intelligence (AI) into mental health care presents both promising opportunities and serious ethical challenges. From diagnostic to therapeutic support tools, AI systems — especially large language models (LLMs) — are progressively being explored to expand access to mental health resources and to improve outcomes. However, their use raises critical concerns around bias, privacy, equity, and effectiveness across diverse populations. A Comment by Nicole Martinez-Martin discusses these challenges, warning that poorly designed AI could worsen disparities in care. Martinez-Martin calls for inclusive design, better data practices, stakeholder engagement, and strong ethical oversight to ensure that mental health AI is safe, fair, and beneficial for all. In a Perspective, Yi-Chieh Lee and colleagues study the potential of LLMs in psychotherapy, by identifying the distinct roles of LLMs in the field, examining associated computational and ethical challenges, and calling for the responsible development and deployment of LLMs in mental health care.

Computational approaches are also reshaping psychiatry when it comes to delivering precision mental health care. In a Viewpoint, a group of authors explore a range of strategies in this context, from leveraging brain-based models to better capture individual neurobiological variation, to developing clinically grounded and interpretable computational tools that align with real-world practice. While some challenges still exist, including limitations of current outcome measures and the need for rigorous model validation, different approaches are suggested by the authors to more accurately reflect the complexity of mental health, such as computational phenotyping and neurodiversity-informed frameworks.

World Mental Health Day is a time to acknowledge the growing need for support for this silent epidemic and the transformative potential of research that explores the mechanisms and treatment options in the field. As computational science continues to unlock new ways of modeling the brain and mind, we hope this Focus can contribute to a larger conversation prioritizing empathy and rigor.