Plenary-03 : Leah Edelstein-Keshet Prize

Kathleen Curtius

Assistant Professor, Department of Medicine

UC San Diego, USA

Abstract:

Cancer initiation and progression is an evolutionary process. Genetic and epigenetic alterations underpin phenotypes that drive natural selection at the cellular level. However, detecting signs of cancer early often requires more than mutation identification alone. Understanding the dynamics is critical to predicting individual cancer risk in patients and intervening effectively. In this talk, I will present mathematical methods that can incorporate multiscale data types (from population-level incidence to patient-level genotypes) to help answer complex questions such as, 'when should we recommend screening/intervention in this at-risk group?' Here we will focus on models for esophageal adenocarcinoma (EAC) and colitis-associated colorectal cancer (CA-CRC). Both EAC and CA-CRC evolve from defined pre-cancerous stages that warrant surveillance, yet current clinical guidelines are suboptimal. In EAC, we constructed stochastic branching process models to determine optimal screening ages and explored patterns of microbial evolution during progression using whole genome sequencing data. In CA-CRC, we derived a cost-effective genomic biomarker for risk of progression and applied artificial intelligence methods to combine accurate epidemiological and clinical data with patient genetic information in a large U.S. healthcare system. Overall, computational models serve as powerful tools for transforming the future of cancer early detection.