Talk at University of Oxford on May 13, 2014
Advances in parallelized 3-D agent-based cancer modeling and digital cell lines
Cancer modeling has been increasingly data-driven in the past several years, leading to improved model hypotheses, more rigorous validation of predictions, and a more broadly-accepted role for mathematical and computational modeling in cancer biology and oncology. However, challenges remain: model calibration remains difficult and hindered by a lack of experimental and simulation data standards. This lack of standards also impedes progress in comparing, cross-validating, and combining models into larger frameworks. Accurate predictions increasingly require 3-D modeling of cells and the microenvironment, but existing toolchains for configuring, simulating, analyzing, and visualizing these systems are fragmented and not designed for ease of use.
We will discuss our early results in calibrating an off-lattice agent-based model of ductal carcinoma in situ (DCIS) to individual patient pathology, and discuss improvements to the model biology. We will demonstrate new advances that allow us to simulate systems of 500,000 or more cells in 8-10 mm3 tissues on high-end workstations. And we will discuss our work to create digital cell lines: a new standard representation of high-content in vitro measurements of microenvironment-dependent cell phenotypic parameters. We will also discuss digital snapshots for sharing simulation state data. Much like biologists can perform experiments on standardized breast cancer cell lines and compare their results, it is our hope that modelers can digital cell lines and snapshots in cancer simulations to help aggregate vital biological data, and to share, combine and improve their simulation results.