10/n
Looking at ‘RCHOP+X’ trials, we used clinical data on each ‘drug X’ to predict the clinical trial results.
Only Pola-R-CHP, and Tucidinostat + R-CHOP, were expected to succeed, and indeed they did
www.asco.org/abstracts-presentations/ABSTRACT451754
www.nejm.org/doi/full/10.1056/NEJMoa2115304

9/n
We calibrated the model to reproduce Progression-Free Survival for the CHOP and RCHOP regimens for Diffuse Large B-Cell Lymphoma.

Simulated tumor population shrinkage agreed well with observed changes in circulating tumor DNA after the first cycle of RCHOP:

8/n
From this ‘bottom-up’ model of tumor heterogeneity, simulating treatment responses in a cohort of patients produces a Kaplan-Meier curve of Progression-Free Survival:

7/n
This extends to combination therapy by using a different dimension of heterogeneity for each drug

This way, patients and cells vary in their sensitivity to different drugs; for example, some patients can be more sensitive to one drug than another, or sensitive to both, etc.

6/n
Extending to patient variability, a group of patients - say in a clinical trial - also have a distribution of drug response phenotypes, with each patient’s cancer containing a range of cellular heterogeneity around the average drug sensitivity of that individual.

5/n
In this model of heterogeneity as a distribution of states, each cycle of chemotherapy progressively shifts the distribution to increasingly drug-resistant states

4/n
Many insightful models of tumor heterogeneity described drug-sensitive and drug-resistant subpopulations.

Based on clone-tracing, we modelled cellular heterogeneity as a distribution of sensitivity phenotypes, reflecting many complex influences on drug response