Developing Predictive Models Using Minimally-Invasive Biomarkers - Part 1/2

Learn how to develop predictive models for cancer treatment using minimally-invasive biomarkers in this conference talk from IPAM's Mathematics of Cancer workshop. Explore the transformation of patient biomarker measurements into dynamic forecasts of disease and treatment response, focusing on clinically actionable prediction rather than retrospective analysis. Discover why minimally- and non-invasive biomarkers like liquid biopsy measures and patient-reported outcomes (PROs) are better suited for longitudinal modeling compared to invasive biomarkers such as tissue biopsies, despite their lower biological resolution. Examine key modeling challenges including noisy measurements, indirect observability of disease state, and inter-patient heterogeneity, along with methods for assessing predictive ability while considering these issues. Follow a detailed demonstration of how PRO data collected biweekly from non-small cell lung cancer patients receiving immunotherapy can be integrated into calibrated and validated predictive models to forecast volumetric disease progression and generate patient-specific predictions. Address practical modeling challenges such as working with noisy and incomplete longitudinal data, balancing biological realism with mathematical tractability, and learn strategies for overcoming these obstacles in predictive oncology models.