Fourteen disorders, five shared genetic roots

Psychiatry has always had a classification problem. The disorders in its manuals are defined by clusters of symptoms, a practical arrangement for diagnosis, but no one has ever been confident that the symptom clusters mark distinct biological diseases. Depression and anxiety overlap in patients, in drugs and, it now appears, in DNA. A study published in Nature on 11 December brings the largest dataset yet to bear on the question, screening the genetic data of more than one million people, and reports that 14 psychiatric conditions sort themselves into five groups, within which most of the genetic risk is shared. Read through DNA, the architecture of mental illness looks less like a list of separate disorders and more like a map of connected regions.

The scale of the analysis is what gives the result its weight. A million people is enough DNA to detect the small individual variants that psychiatric genetics has learned to expect: most risk alleles shift the odds by a few per cent each, and only enormous samples can separate real signals from noise. With that power, the team asked a structural question: not whether a given gene raises the risk of schizophrenia, but whether the same genetic signals travel together across conditions. The answer is that they do, and the travelling is organised. Depression, anxiety and schizophrenia, along with a set of other conditions, sit in networks where a large fraction of the heritable risk is shared rather than condition-specific.

The five-group structure is the headline, but the useful reading is about relationships rather than boundaries. Conditions that share genetic architecture are likely to share biology, which means shared drug targets and the possibility of treatments that work across several conditions at once. For a field where drug development is slow and expensive, a genetic map that points to shared mechanisms is a shortcut to the most plausible places to look. The findings could also reshape diagnosis: if the genetic evidence groups conditions that the manuals keep apart, then the manuals, built on symptoms, may be drawing lines where biology has borders of a different kind.

The study is careful about what a shared signal does not mean. Shared genetic risk is not the same as shared cause, and the environment, development and experience still determine which of the predisposed people develop which condition, or any condition at all. Nor does the map dissolve the categories into one: the five groups are distinct, with their own signals, and the paper reports both the sharing and the differences. The result is not that psychiatric diagnoses are meaningless; it is that they are connected in ways the symptom-based system could not see.

The significance is in what it promises for treatment. Most psychiatric drugs were discovered by chance or by analogy, and their targets are a fraction of the genes that the new data implicate. A genetic architecture this structured suggests that some of the shared mechanisms will be druggable, and that a therapy aimed at one group’s common biology could serve many conditions within it. The manual will not be rewritten overnight, and the clinic will not change next year. But the map of mental illness, drawn for the first time at this scale, is a map of where to dig, and digging has always been cheaper than guessing.