Glioblastoma is noted for its ability to aggressively invade the brain tissue beyond what may be visualized clinically. This is due in part to the unique architecture of the brain, and also to the aggressiveness of the invading glioma cells. To a good degree, the bulk growth and extent of invasion of each patient's tumor can be estimated from
imaging, and the future growth dynamics can be predicted. But the response following treatment is less predictable, because the same growth parameters can be present with different heterogeneous tumor compositions, which could lead to different post-treatment responses. It is well accepted that brain tumors are heterogeneous genotypically, phenotypically, and spatially. Observations made from clinical imaging only gets to the level of bulk properties, so to get a better idea of the behavior of individual cells, we use an experimental rat model that simulates glioblastoma and allows for acquisition of single-cell, in situ, temporal migration and proliferation data. Tightly intertwined within these observed data is the net effect of both the possible phenotypic range that an individual cell may exhibit and the context-dependent manifestation of that phenotype due to a particular environment at a particular time. For tumor growth, this external and internal heterogeneity is present and impossible to separate. Many different solutions can be used to get similar bulk growth rates, but the individual cell heterogeneity becomes particularly significant when the environment changes through applied treatment.

In order to gain insight on the source of observed heterogeneity, we build an agent based model that examines each source separately, the autonomous individual and the external environment, while keeping the other constant. The distributions of proliferation and
migration parameters from the experimental data are used to calibrate the cell based model. We simulate how these differences in phenotypes with variation in the environment (growth factor gradients and competition for space) affect the evolution of heterogeneity in a tumor both spatially and temporally and lead to both bulk tumor growth differences and variation in the distributions of individual cell observables.