Attention-deficit hyperactivity disorder (ADHD) is one of the most common neurodevelopmental disorders. Red flags signaling the condition are often picked up during childhood when children experience problems paying attention or controlling impulsive behaviors.
Diagnosing ADHD is not straightforward. Without a single test to positively identify the condition, children often have to endure a battery of tests, and physicians don’t always get the diagnosis right the first time.
Scientists say the answers may be in the genes. Distinct duplications or deletions in the genome, known as copy number variation (CNV), are associated with neurological disorders, including ADHD. Based on previous studies, ADHD is more likely caused or impacted by a network of underlying pathways as opposed to a single gene variant. Artificial intelligence (AI) has the potential to detect these pathways. However, these models have yet to be established in the context of ADHD.
Now, a team of genomics experts from the Children’s Hospital of Philadelphia report that machine learning could help boost the accuracy of current ADHD diagnostic methodologies. Deep-learning methods average a 78 percent accuracy, a step up from conventional techniques that average 50 percent.
The team collected whole-genome sequencing data from over 500 African Americans (a group often underrepresented in existing genetic databases). They then applied multiple layer deep learning algorithms to capture the genetic susceptibility to the disease and the molecular pathways influencing ADHD.
The study was led by Yichun Liu from the Center for Applied Genomics (CAG). Liu said that AI deep learning paired with advanced sequencing techniques stand to benefit ADHD patients greatly. “Combining both together forms a powerful tool for the prediction and diagnosis of complex mental disorders, such as ADHD, especially for toddlers who are too young to be diagnosed,” explained Liu.