AI-Driven Biomarker Discovery from Wearables in Neuropsychiatric Research

Mark Gerstein, PhD, discussed the application of artificial intelligence (AI) and wearable device data to the identification of neuropsychiatric biomarkers, with a focus on attention-deficit hyperactivity disorder (ADHD) and anxiety disorders.

Gerstein described a paper published recently in which his group developed objective biomarkers derived from wearable devices for various neuropsychiatric conditions. A central methodological innovation was repurposing genome-wide association study machinery, which is traditionally used to identify associations between genetic variants and phenotypic traits, to instead work with wearable-derived biomarker data. The study drew heavily on the Adolescent Brain Cognitive Development study, which Gerstein described as a rigorously assembled dataset incorporating genetics, clinical phenotyping, wearable data, and neuroimaging.¹ He emphasized that access to such well-curated, multimodal datasets represented the single most critical prerequisite for this class of research.

Gerstein articulated a clear rationale for the indispensability of AI in this domain. Wearable devices generate dense, multi-channel streams of time-series data that human observers are not naturally equipped to interpret.² He described AI's core strength as pattern recognition in precisely those data domains that fall outside human intuitive expertise: "the AI is really important to take that information and use it to predict or to identify the disorder, and what in the process it finds as key features that it's honing in on" to produce actionable biomarkers. He contrasted this with speech analysis, noting that while large language models had demonstrated impressive speech processing, clinicians already possess strong intuitive competence in that modality. Gerstein argued that AI's value is therefore greatest "in parsing data that we just do not know what to do with, like these streams of information from wearable devices."

Looking ahead, Gerstein expressed enthusiasm about the expanding wearable ecosystem, including smart eyeglasses, earbuds, and rings, as a growing source of passively collected behavioral and physiological data that could eventually support longitudinal psychiatric monitoring and disease detection.

Dr Gerstein is the Albert L. Williams Professor of Biomedical Informatics at Yale, affiliated with the departments of molecular biophysics and biochemistry and statistics and data science, along with the departments of computer science, biomedical informatics, and data science.

References

1. Adolescent Brain Cognitive Development Study. National Institute on Drug Abuse. November 17, 2025. Accessed June 4, 2026. https://nida.nih.gov/research-topics/adolescent-brain/longitudinal-study-adolescent-brain-cognitive-development-abcd-study

2. Vijayan V, Connolly JP, Condell J, et al. Review of wearable devices and data collection considerations for connected health. Sensors (Basel). 2021;21(16):5589.