Navigating Sparse and Noisy Data: Machine Learning Approaches for Ovarian Cancer

High-grade serous ovarian cancer (HGSC) remains one of the most heterogeneous and lethal malignancies, yet our ability to capture and interpret this complexity is limited by the technologies we use to study it. Bulk sequencing provides population-scale coverage but lacks cellular resolution, while single-cell and spatial profiling reveal fine-grained biology but remain constrained to small cohorts. Our lab is developing machine learning frameworks that bridge these worlds, integrating high-resolution but small-scale data with large-scale bulk datasets, to uncover clinically relevant and generalizable tumor features that are otherwise hidden.

We are particularly focused on understanding HGSC transcriptomic subtypes, which differ in survival outcomes but remain poorly characterized across tumor compositions and patient populations. By combining multimodal data integration with interpretable models, we aim to reveal whether these subtypes arise from tumor-intrinsic or compositional differences.

Dr. Davidson is an Asst. Professor at CU Anschutz. Her research program utilizes machine learning to transform the diagnosis & treatment of high-grade serous carcinoma (HGSC) of tubo-ovarian origin. The goal is to develop computational frameworks that uncover biologically meaningful patterns from large, multi-modal datasets, enabling new insights into disease mechanisms & accelerating the translation of these discoveries into improved patient care.