AIMI Grand Rounds: Genetic Diagnosis and Discovery Enabled by Large Language Models - Konstantina Stankovic, MD, PhD

Speaker:

Konstantina Stankovic, MD, PhD: Bertarelli Foundation Professor and Chair, Stanford University 

Bio: Konstantina “Tina” Stankovic, MD, PhD, FACS is the Bertarelli Foundation Professor and Chair of the Department of Otolaryngology – Head & Neck Surgery at Stanford University School of Medicine, and professor, by courtesy, of Neurosurgery. She is a Harvard-trained ear and skull-base surgeon and a Massachusetts Institute of Technology-trained auditory neuroscientist. She blends her surgical expertise with training in physics, molecular biology, auditory neuroscience, and systems electrophysiology to devise novel solutions tailored to the unmet needs of those with hearing loss. She has initiated and led cross-departmental, national, and international collaborations to develop and deploy novel molecular diagnostics and therapeutics for hearing loss and otologic diseases in general while educating the next generation of leaders in surgery and science. She is an elected member of the National Academy of Medicine.

Abstract: Artificial intelligence (AI) has been used in many areas of medicine, and large language models (LLMs) have shown potential utility for various clinical applications. However, to determine if LLMs can accelerate the pace of genetic diagnosis and discovery, we examined whether recently developed LLMs (Med-PaLM 2 and Gemini) could assist in solving four types of genetic problems, which had sequentially increased complexity. First, in response to free-text input, Med-PaLM 2 correctly identified murine genes with experimentally verified causative genetic factors for six previously studied murine models for biomedical traits. Second, Med-PaLM 2 identified a novel causative murine genetic factor for spontaneous hearing loss that was validated using knock-in mice. Third, we developed a retrieval and grounding pipeline that enabled Gemini 2.5 Pro to analyze large lists of genes with genetic variants, which were identified in the genomic sequences of 20 human subjects with hearing loss; and it identified causative genetic factors for 80% of these subjects. Fourth, we developed a genetic analysis pipeline, which enabled the modified Gemini 2.5 Pro without any task-specific fine-tuning to identify causative genetic factors for six subjects with rare genetic diseases that required 14 to 34 different terms to describe their multi-faceted symptom complex. These results demonstrate that an AI pipeline can facilitate genetic diagnosis and discovery in mice and humans.