A new study, published in the journal Molecular Psychiatry, has developed a deep learning model that is able to genetically differentiate patients with mental disorders from control patients.
The mental health crisis
Mental disorders are a leading global health concern, with millions of people worldwide suffering from mental health conditions. In 2019, in the United States alone, 13.1 million adults experienced serious mental illness. In addition, suicide is currently the second leading cause of death among people aged 10 to 34.
Unfortunately, many health services are not yet equipped to deal with mental disorders. The average delay between the onset of mental disorder symptoms and treatment is 11 years and many individuals get misdiagnosed. For patients from minority ethnic backgrounds, barriers in diagnosis and treatment are even harder to overcome. This is partly due to the historic under-representation of minority ethnic populations in studies addressing the underlying genetic causes of mental disorders.
Mental health and genetics
Previous research has shown that a variety of mental disorders have strong associations with structural genomic variation. Some genetic variants have even shown potential as future drug targets. In addition, deep learning algorithms have previously been used to successfully diagnose mental disorders, such as bipolar disorder.
In the current study, researchers decided to explore mental health, genetics and deep learning further, whilst also addressing the lack of minority ethnic populations in such studies.
Building a deep learning model
Firstly, the team performed whole genome sequencing on blood samples from 4,179 African American patients. Out of this cohort, 1,384 had been diagnosed with at least one mental health condition.
The study focused on eight common mental disorders. These were: ADHD, depression, anxiety, autism spectrum disorder, intellectual disabilities, speech/language disorder, delays in development and oppositional defiant disorder (ODD).
The team used the collected data to train a deep learning model with the aim of predicting and diagnosing mental disorders.
Identifying patients with mental disorders
Upon testing, the researchers found that their deep learning tool had approximately 65% accuracy in differentiating patients with mental disorders from control patients. However, when the researchers attempted to predict the diagnosis for patients with multiple disorders, accuracy was much lower. The model’s prediction only matched the patient’s known conditions 7-10% of the time. Although this sounds low, the researchers point out that random guess accuracy is around 0.4%, so the deep learning model is not relying on randomness.
Getting a diagnosis for a mental health disorder is still a big global issue. The deep learning tool developed in this study has the ability to differentiate patients with mental disorders from controls. This may help with future diagnostics. Further work will be required to increase the accuracy of the model in diagnosing patients with multiple disorders. However, this study is an excellent starting point for future research.
Senior author Hakon Hakonarson said:
“Most studies focus only on one disease, and minority populations have been very under-represented in existing studies that utilise machine learning to study mental disorders. We wanted to test this deep learning model in an African American population to see whether it could accurately differentiate mental disorder patients from healthy controls, and whether we could correctly label the types of disorders, especially in patients with multiple disorders.”
Picture credit: Canva