Mental health conditions can run in families. Studies of identical twins, for example, show higher chance of shared disorders than siblings and cousins (Polderman TJC et al., 2015). This has led researchers to look for specific genes involved in psychiatric conditions. The hope is that identifying these genes and the proteins they produce could lead to more targeted and effective treatments.
One major method used is the genome-wide association study (GWAS). In GWAS, large numbers of people provide DNA samples and their individual genotypes (the exact genetic code at many sites in their genome) is measured to identify changes in the “letters” of their DNA, known as single nucleotide polymorphisms (SNPs). These are then compared with a particular characteristic (called a phenotype), such as having symptoms of depression/psychosis, to determine which SNPs might be related to the phenotype.
A strength of GWAS is it examines the whole genome at once, allowing a broad look at gene associations. Modern GWAS include data from thousands of people and are often widely accessible by researchers (e.g. UK Biobank), improving transparency and reproducibility. Findings from GWAS studies can be helpful in informing our understanding of many phenotypes, including identifying treatment targets for many diseases.
However, GWAS have weaknesses. They primarily detect common SNPs, missing rarer genetic changes/variants. Many studies also rely on data from people with a European ancestry, limiting applicability across diverse populations. A related problem is “population stratification”, where differences in genotype and phenotype frequencies between populations (possibly due to chance or environmental factors) may create false associations, even in the absence of a real causal relationship.
In psychiatry, GWAS have uncovered genes involved in many disorders. The latest GWAS for depression (Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium, 2025) included data from over 5,000,000 people and found 308 gene associations with depression. These genes were more active in emotion-related brain regions, like the amygdala, and were linked to targets of antidepressants.
Medications like antidepressants can be incredibly helpful for some patients. However, we do not really understand how the action of medications at the level of individual proteins or cells relate to their subjective impacts on how people think and feel. Genetic studies might help shed further light on this and this was the aim of the paper we are focusing on in this blog – Arnatkeviciute and colleagues (2025).
Genetic studies like GWAS help identify mental health risk genes, but challenges like limited diversity and false associations remain.
Methods
The goal of the study was to see how much the genes linked by GWAS to mental health problems overlapped with those targeted by psychiatric medications. The authors used openly available datasets and focused on four psychiatric conditions: attention-deficit/hyperactivity disorder (ADHD), bipolar disorder, schizophrenia, and major depressive disorder. They also included type 2 diabetes, heart failure, rheumatoid arthritis, and inflammatory bowel disease as non-psychiatric comparisons.
They used information from psychiatric disorder GWAS, gene function, how proteins produced by genes interact and drug-protein interactions from DrugBank. Their dataset included 2,232 genes listed as interacting with approved medications (14 for ADHD, 29 for schizophrenia, 48 for depression and 22 for bipolar disorder).
Each gene was given two scores: A drug score based on whether its protein product was targeted by the drugs, and a GWAS score based on four strategies:
- Single nucleotide variant (SNV) position: How close a gene is to specific spots in the DNA that have been linked to mental health conditions. Higher scores mean the gene is near more of these linked DNA changes.
- Protein-protein interaction (PPI) network: Whether the protein made by a gene interacts with other proteins that are linked to mental health conditions. This was a generalisation of strategy (1).
- Brain expression quantitative trait loci (eQTLs): Whether changes in DNA affect how much a gene is turned on or off in the brain. Higher scores indicate that the gene’s activity in the brain is more affected by these DNA changes.
- Spatial gene expression: How much the pattern of where a gene is active in the brain overlaps with the pattern of brain activity for genes linked to mental health conditions. Higher scores mean the gene’s brain activity pattern closely matches that of the genes identified by GWAS for psychiatric disorders.
The drug and GWAS scores were compared to measure gene overlap and were then checked against random drug sets to see if the overlap was greater than expected by chance.
Results
Using their PPI network method, the authors found that there was statistically significant overlap between genes linked to mental health conditions by GWAS and genes targeted by medications for bipolar disorder and type 2 diabetes. For all other mental or physical health conditions studied, no meaningful overlap was found for any of the methods used. This suggests that some disorders may be genetically linked to specific treatments, while others may not be.
The authors then explored the functions of genes identified by GWAS for psychiatric disorders and found that bipolar disorder-associated genes were especially linked to neurotransmission at synapses. The authors suggest that this may explain why these genes overlap more with the targets of bipolar disorder treatments.
Next, the authors looked at the association between bipolar-GWAS genes and specific types of bipolar disorder medications. Again, using the PPI method, they found:
- A strong overlap with anticonvulsant medications, a type of mood stabiliser.
- Some overlap with antipsychotic medications.
In addition, an analysis was performed to see if there was any overlap between the genes linked to the assessed psychiatric disorders and the medications used to treat them – no new significant associations were found.
Finally, the authors performed a series of sensitivity checks to explore whether their choice of analysis impacted their findings. When they widened their PPI network to include more proteins they found some evidence that genes linked to major depression also overlapped with medications used to treat depression (in addition to the overlaps already seen with bipolar disorder and type 2 diabetes).
Bipolar disorder and type 2 diabetes showed a significant genetic overlap with their treatments, while depression showed some overlap, but no other conditions exhibited meaningful genetic connections.
Conclusions
The authors conclude that there was little overlap between the genes linked to the risk of mental health conditions and the targets of pharmacological treatments — with the exception of bipolar disorder, which they argue has a stronger genetic connection to synaptic proteins. They also showed that, for non-psychiatric conditions, only type 2 diabetes showed a pattern of genetic overlap between the condition and its treatments.
Overall, few conditions showed significant genetic overlap with their treatments, with bipolar disorder and type 2 diabetes being notable exceptions.
Strengths and limitations
Strengths
The authors have demonstrated excellent technical skills by bringing together multiple complex datasets to carry out their analysis, which is no easy feat. They used a structured reporting guideline (Strengthening the Reporting of Genetic Association Studies, STREGA), which provides a consistent format for papers and may help authors make sure they are giving consistent information to readers.
Limitations
Although this study asks an interesting question, the approach used may not be the best way to answer it and the implications for clinical practice remain unclear. It is not immediately clear that the genetics of a disorder should necessarily overlap with its treatment, or that it would be a problem if they did not. Their results are also contradicted by the most recent depression GWAS, which found that depression-associated genes are enriched for antidepressant targets (Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium, 2025).
The causal processes behind health problems are complex and involve the interplay of a wide range of factors including genetic, environmental and social exposures. It is notable that although mental disorders are heritable, SNP heritability is low (e.g. the most recent estimate for major depression in European-ancestry individuals is 5.8%; Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium, 2025). The contribution of a wide range of environmental factors (e.g., adverse childhood experiences, substance/alcohol use), are not captured within GWAS. The effectiveness of a treatment is based on its clinical evidence, not necessarily its genetic link to a condition’s process.
In a simplistic example, it is well established that smoking is a key cause of lung cancer. However, while there are genes that increase one’s risk of smoking (Gorman et al. 2024), the genes themselves do not directly increase the risk of a lung tumour – it’s the act of smoking itself that does. Consequently, treatments targeting smoking-associated genes might influence smoking behaviour, but would not be useful for the treatment of lung cancer. Similarly, in Type 1 Diabetes the “problem” might be high blood sugar, but the cause is the destruction of insulin producing beta cells in the pancreas by the body’s own immune system. Genetic studies in Type 1 Diabetes implicate immune function (Grant et al., 2020), but the typical insulin treatment does not address the cause of diabetes, only the consequences. Nevertheless, insulin is the most appropriate treatment.
It’s striking that the authors emphasise how Type 2 diabetes showed more overlap with its treatment targets than several psychiatric disorders, but the authors later note that no overlap was found for other non-psychiatric conditions (heart failure, rheumatoid arthritis or inflammatory bowel disease). This suggests that mental health disorders may be similar to physical health disorders showing little overlap between GWAS and medication-associated genes.
Another limitation is the author’s choice of medications. “Bipolar depression” was excluded from their search of bipolar disorder drugs, but they included multiple antidepressants used in the treatment of depression in bipolar disorder (e.g. bupropion, fluoxetine). Similarly, the schizophrenia treatments include antidepressants with no antipsychotic efficacy, and the major depression medication list includes antidepressants, stimulants, antipsychotics and mood stabilisers.
Lithium, the gold standard treatment for bipolar affective disorder and a treatment for major depression, was inconsistently handled. Three options for lithium were included in the bipolar dataset, but only one in the major depression dataset.
In psychiatry, comorbidity is common, as is using medications from one class to treat a variety of disorders. In this way the author’s choices reflect clinical practice. However, it makes it much harder for the authors to conclude they are measuring distinct medication sets for distinct disorders.
Though the authors do split bipolar disorder treatments by class, they did not do so for other conditions. A more useful approach would be to run their analysis by medication class or using a more modern drug target nomenclature, like Nbn2 (https://nbn2r.com/) or the groupings proposed by McCutcheon et al. (2023).
This paper provides valuable insights into the genetic overlap between disorders and treatments, but is limited by inconsistencies in medication selection and the broader applicability of the findings.
Implications for practice
The findings that anticonvulsants are genetically linked to bipolar disorder raises interesting and highly clinically relevant questions about their role in treating this disorder. In Nick’s own clinical experience, anticonvulsants have some specific roles but are not first-line agents. According to the NICE guidelines for bipolar disorder, lithium is the gold standard for long-term treatment, alongside antipsychotics. Valproate (in its various forms) is a second- or third-line treatment option for mania, while lamotrigine is used for treating depression in bipolar disorder.
Recently, valproate has been increasingly restricted due to concerns about its teratogenic effects (causing non-heritable developmental defects) in both women and men (see gov.uk news story here). Lithium, while effective, requires regular blood test monitoring and can affect thyroid and kidney function. Antipsychotics can cause weight gain, sedation and movement problems, among other side effects. Valproate was a good alternative option, but it is now less readily available.
A potential direction for future research could be the development of drugs with a similar mechanism of action to valproate, but without its potential adverse effects. Such advances could offer safer, more accessible treatment options for bipolar disorder.
The genetic link between anticonvulsants and bipolar disorder highlights potential new treatment directions, particularly as valproate becomes less accessible due to safety concerns.
Statement of interests
Nick wrote the first draft of this blog and has no personal or professional link to this study or its authors. Eimear is a coordinator for the Mental Elf and worked on the second draft on the blog. She has no conflicts of interest to declare.
Links
Primary paper
Arnatkeviciute A, Fornito A, Tong J, Pang K, Fulcher BD, Bellgrove MA. Linking Genome-Wide Association Studies to Pharmacological Treatments for Psychiatric Disorders. JAMA Psychiatry. 2025 Feb 1;82(2):151-160. doi: 10.1001/jamapsychiatry.2024.3846. PMID: 39661350; PMCID: PMC11800018.
Other references
Gorman BR, Ji SG, Francis M, Sendamarai AK, Shi Y, Devineni P, Saxena U, Partan E, DeVito AK, Byun J, Han Y, Xiao X, Sin DD, Timens W, Moser J, Muralidhar S, Ramoni R, Hung RJ, McKay JD, Bossé Y, Sun R, Amos CI; VA Million Veteran Program; Pyarajan S. Multi-ancestry GWAS meta-analyses of lung cancer reveal susceptibility loci and elucidate smoking-independent genetic risk. Nat Commun. 2024 Oct 4;15(1):8629. doi: 10.1038/s41467-024-52129-4. PMID: 39366959; PMCID: PMC11452618.
Grant SFA, Wells AD, Rich SS. Next steps in the identification of gene targets for type 1 diabetes. Diabetologia. 2020 Nov;63(11):2260-2269. doi: 10.1007/s00125-020-05248-8. Epub 2020 Aug 14. PMID: 32797243; PMCID: PMC7527360.
Major Depressive Disorder Working Group of the Psychiatric Genomics Consortium. Trans-ancestry genome-wide study of depression identifies 697 associations implicating cell types and pharmacotherapies. Cell. 2025 Feb 15;188(3): 640–652. https://doi.org/10.1016/j.cell.2024.12.002
McCutcheon RA, Harrison PJ, Howes OD, McGuire PK, Taylor DM, Pillinger T. Data-Driven Taxonomy for Antipsychotic Medication: A New Classification System. Biol Psychiatry. 2023 Oct 1;94(7):561-568. doi: 10.1016/j.biopsych.2023.04.004. Epub 2023 Apr 14. PMID: 37061079; PMCID: PMC10914668.
Polderman TJC, Benyamin B, de Leeuw CA, Sullivan PF, van Bochoven A, Visscher PM, & Posthuma D. Meta-analysis of the heritability of human traits based on fifty years of twin studies. Nature genetics. 2025 Jul; 47(7), 702–709. https://doi.org/10.1038/ng.3285
