Careers and Collaboration Speaker Bios 2025
Emily Alsentzer, Ph.D.
Assistant Professor in Biomedical Data Science and, by courtesy, Computer Science
Title: Knowledge-Enhanced Foundation Models for Healthcare
Abstract: Foundation models promise significant advancements in biomedicine, but current large language models often fail to encode the structured knowledge critical for biomedical applications. In this talk, I will discuss our work building domain-specific foundation models enriched with biomedical knowledge. I will introduce a few-shot learning approach that leverages external knowledge in biomedical knowledge graphs to diagnose patients with rare diseases. To infuse external knowledge, we frame diagnosis as a subgraph prediction task, training a graph neural network to represent a patient’s phenotypic and genetic data in relation to a knowledge graph of known phenotype, gene, and disease associations. To overcome the challenges of data scarcity, we developed a framework for simulating realistic rare disease patients, demonstrating strong alignment with real-world patient presentations. We leverage these advances for multi-faceted diagnosis of patients in the Undiagnosed Diseases Network, performing causal gene discovery, retrieving “patients-like-me”, and providing interpretable characterizations of novel disease presentations. This work demonstrates the benefit of incorporating medical knowledge for model generalizability and illustrates the potential of deep learning to accelerate molecular diagnosis for rare disease patients.
Carlos Bustamante, Ph.D.
Adjunct Professor of Biomedical Data Science, of Genetics and, by courtesy, of Biology
Title: Enabling Precision Health at Scale for All
Abstract: (coming soon)
Roxana Daneshjou, MD, Ph.D.
Assistant Professor of Biomedical Data Science; Assistant Professor, Dermatology
Title: Generative AI for healthcare
Abstract: The healthcare system is fundamentally broken. Could generative AI improve the physician and patient experience?
Stephen Montgomery, Ph.D.
Director of Admissions; Professor of Pathology, of Genetics, of Biomedical Data Science and, by courtesy, of Computer Science
Title: Using multi-omics to study undiagnosed rare diseases
Abstract: As many as 50% of people with a rare disease do not receive a clinical diagnosis. I will discuss activities within the NHGRI GREGoR Consortium and GREGoR Stanford Site that are combining multi-omics with novel computational methods to provide new diagnoses.
Cynthia Xinran Li, Ph.D. Candidate in Computational Mathematics and Engineering (ICME)
Title: (to come)
Abstract: The probability of cancer varies across different locations within the prostate. Information about how cancer is distributed spatially may improve biopsy templates and cancer diagnosis and benefit AI model performance. We quantified the spatial heterogeneity of prostate cancer by gathering cancer locations from 2077 prostate MRIs and constructed cancer atlases. Using such atlases as priors, we enhanced the diffusion-based model’s performance in finding cancer.
Dennis Wall, Ph.D.
PhD – Director of Graduate Studies; Professor of Pediatrics (Systems Medicine), of Biomedical Data Science and, by courtesy, of Psychiatry and Behavioral Sciences
Title: An adaptive digital game for healthy child development
Abstract: I’ll treat this talk as a short pitch to support commercialization of a system developed in my lab. I’ll cover the unmet need, market size, approach, and the path to commercialization.