Conference Agenda

Cancer immunotherapy is an important application area for mathematical modeling. Current modeling studies have a range of ambitious goals from dose optimization to creating “digital twins” of individual cancer patients for treatment response prediction. Here we focus on a humbler, but nonetheless important, goal: aiding with the planning and design of clinical trials. Cancer immunotherapy trials can be hard to design due to heterogeneous and time-varying treatment effects. While clinical statisticians already use computer simulations, these rarely integrate explicit pathophysiological mechanisms, such as cancer-immune interactions, to specifically adapt the design to the treatment. Encouraged by rapid progress in mathematical modeling, we here propose an “in-silico-first” approach–already common in industry–where doctors, statisticians, and modelers build knowledge-based mathematical models to examine and refine the statistical design of clinical trials. We will discuss some experiences with such in silico trial designs in the cancer immunotherapy field. We wil also, reflect on how these approaches contrast to using causal diagrams for similar purposes, and offer some opinions on what we feel are advantages and disadvantages of each approach.

Cancer results from a stochastic evolutionary process characterized by the accumulation of mutations that are responsible for tumor initiation, progression, immune escape, and drug resistance, as well as mutations with no effect on the phenotype. Mathematical modeling, combined with clinical, sequencing and epidemiological data, can be used to describe the dynamics of tumor cell populations and to obtain insights into the hidden evolutionary processes leading to cancer. I will present recent approaches for quantifying the evolutionary dynamics of cancer in patients, and their implications for deciphering cancer heterogeneity and response to therapy.

Directed acyclic graphs (DAGs) play a large role in the modern approach to causal inference. DAGs describe the relationship between measurements taken at various discrete times including the effect of interventions. The causal mechanisms, on the other hand, would naturally be assumed to be a continuous process operating over time in a cause–effect fashion. At the same time DAGs are less suited to provide insight into the actual effects of a given intervention on a desired outcome. The reason being that DAGs are really just an encoding of direction and absence of causal effect. It does not naturally express magnitudes or even directions. Finally, there are multiple misconceptions on if/how feedback can be expressed in DAGs. In this talk I will present how classic DAGs can be combined with systems dynamic thinking and simulations to increase their usability. It will also explore how the underlying principles in causal inference thinking (eg keep the scientific question in focus as well as clearly defined) can also strengthen simulation based methos.