About the Speaker
Na Cai is an Assistant Professor in the Department of Biosystems Science and Engineering (D-BSSE) at ETH Zurich. Dr. Na Cai’s research focuses on understanding how genetic variants influence the risk of neuropsychiatric diseases, either directly or in combination with physiological and environmental factors. Her group also develops and applies machine learning methods to electronic health records.
She completed a postdoc at the Wellcome Sanger Institute and EMBL-EBI, and subsequently led her own research group at Helmholtz Munich starting in 2019. In February 2025, her group moved to D-BSSE at ETH Zurich, where she now serves as Assistant Professor in Computational Medical Genomics.
Talk Abstract
Genetics research is transforming with the growing availability of population-scale biobanks and electronic health records (EHRs). Not only do these datasets provide unprecedented sample sizes, they offer different data modalities than those traditionally used in complex trait and disease investigations. In particular, the ICD10 code or Phecode systems are used to identify individuals with a disease in EHRs, in replacement of rich clinician assessments in traditional case-control cohorts as well as self-administered questionnaires in volunteer-based biobanks. While studies using EHR codes identified biases, cautioned against them, and developed means to correct for them in prediction tasks (for future disease episodes or outcomes), less attention have been paid to removing these biases and refining the EHR codes such that they are more useful for genetic studies. In this talk, I would present our most recent work on refining EHR codes using deep-learning methods for 9 complex diseases, each with its own biases. I would demonstrate how disease-specific information from self-administered questionnaires or laboratory values can improve EHR code refinement, and how best to select individuals to obtain this additional information in call-back studies.
We are looking forward to the talk!