The authors argue that the complexity of mental illness requires computational approaches capable of linking processes across multiple levels, from brain function to behaviour and environmental influences. They describe two complementary approaches: data-driven methods, which use machine learning to improve diagnosis and prediction, and theory-driven models, which seek to explain the mechanisms underlying psychiatric symptoms and behaviour. Computational psychiatry offers new opportunities for identifying clinically meaningful subgroups, predicting treatment outcomes, and refining psychiatric classification systems. Overall, the paper positions computational psychiatry as a promising bridge between advances in neuroscience and real-world clinical applications.