An AI-driven study published in Nature Genetics has identified 766 genes associated with schizophrenia, 641 of which had never surfaced in previous transcriptomic analyses. The work drew on genetic data from more than 102,000 people and brain tissue samples from six regions in hundreds of donors, producing one of the most detailed maps yet of the disorder's genetic architecture. Schizophrenia affects roughly 23 million people worldwide, or about one in every 345, according to the World Health Organization.
The scale matters because schizophrenia does not behave like a single-mutation disease. Hundreds of genetic variants each contribute small effects across different brain processes — neural development, synaptic communication, the organization of brain connections — and no single variant explains the illness. Untangling that network requires computational models capable of reconstructing coordinated activity across thousands of genes simultaneously, work that was not tractable at this resolution a decade ago.
That is where AI enters the picture. The team used machine-learning models to detect long-range genetic regulatory signals — cases where a variant in one part of the genome influences the expression of a gene elsewhere. Many of the 641 newly implicated genes were surfaced through exactly these signals, evidence that the genes involved in schizophrenia function as an interconnected network rather than as isolated hits. Standard association studies, which look for direct statistical links between variants and disease, would have missed them.
Key facts
- 01A Nature Genetics study identified 766 genes tied to schizophrenia, 641 of them not seen in prior transcriptomic analyses.
- 02The team analyzed genetic data from more than 102,000 people and brain tissue from six regions across hundreds of donors.
- 03Schizophrenia affects roughly 23 million people worldwide, about one in every 345, according to the World Health Organization.
- 04Researchers from the Lieber Institute for Brain Development and the University of Bari led the multi-country collaboration.
- 05The AI models reconstructed coordinated activity across thousands of genes to reveal long-range regulatory signals.
Researchers from the Lieber Institute for Brain Development, the University of Bari, and dozens of psychiatric centers across multiple countries participated in the project. The brain tissue samples spanned six distinct regions, allowing the models to compare gene expression patterns across the cortex, subcortical structures, and other areas implicated in psychiatric disease. That regional breadth is a departure from earlier work, which often relied on tissue from a single brain area and consequently missed variation that only appears when the whole organ is considered together.
The researchers describe the finding as turning on the lights in an entire neighborhood. Until now, they could observe a few lit houses — the handful of high-confidence risk genes identified in prior studies — but now they can make out a much larger portion of the disease's genetic map. Rather than acting separately, the variants appear to coordinate and collectively contribute to the risk of developing schizophrenia.
The clinical implication is that drug discovery for schizophrenia has been constrained by a shortage of validated biological targets. Current antipsychotics largely modulate dopamine and serotonin pathways that have been known for decades, and response rates are inconsistent. A larger, better-mapped set of causal genes gives pharmaceutical researchers more starting points for both small-molecule drugs and, potentially, gene-directed therapies aimed at the regulatory networks the study uncovered.
Family history is known to increase schizophrenia risk without determining it. Some people with close relatives who have the condition never develop it; others are diagnosed with no known family history at all. The polygenic, network-based picture emerging from this study helps explain that pattern: risk is distributed across many small-effect variants that combine differently in each individual, and inheritance of any single variant is not decisive.
The disease alters perception of reality and typically manifests through hallucinations and delusions. It can also produce social isolation, lack of motivation, attention problems, memory difficulties, and thought disorders. For many researchers, this diversity of symptoms reflects the complexity of the biological mechanisms underlying the disorder — there does not appear to be a single gene responsible, but rather an extensive network of interacting processes.
The limits of the work should be stated plainly. Identifying 766 associated genes is not the same as identifying 766 drug targets; the majority will require years of functional validation before any therapeutic hypothesis can be tested. The models also identify statistical associations and regulatory relationships, not causation, and specialists caution that they still do not know how the numerous biological factors contributing to schizophrenia actually interact in a living brain. Reproducing the results in more diverse population cohorts will also be necessary.
The broader signal for the AI market is that computational biology is moving from pattern-matching on isolated datasets to reconstructing full regulatory networks across tissues, populations, and disease states. The bottleneck in psychiatric drug discovery has never been chemistry; it has been the absence of validated biological targets grounded in human data. Models capable of surfacing 641 previously invisible genes from a single study change the economics of that pipeline, and they explain why pharma is spending aggressively on the same class of infrastructure that trains frontier language models. The next round of progress in complex disease will look less like a new molecule and more like a better map.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
