# Novel Mixed Cancer-Cell Models Designed to Capture Inter-Patient Tumor Heterogeneity for Accurate Evaluation of Drug Combinations

## TL;DR

This paper introduces new mixed cancer-cell models that better reflect patient tumor diversity, improving the evaluation of drug combinations for prostate cancer.

## Contribution

The novel contribution is the development of mixed-cell models that capture inter-patient tumor heterogeneity for more accurate drug combination testing.

## Key findings

- Mixed-cell models accurately reflected clinical drug responses in castration-resistant prostate cancer.
- Previously undetected drug combinations showed efficacy in the mixed-cell models.
- Traditional models failed to detect the efficacy of these combinations.

## Abstract

Disease heterogeneity across a diverse patient cohort poses challenges to cancer drug development due to inter-patient variability in treatment responses. However, current preclinical models fail to depict inter-patient tumor heterogeneity, leading to a high failure rate when translating preclinical leads into clinical successes. We integrated the expression profiles of prostate cancer (PC) lines and castration-resistant PC (CRPC) patient tumors to identify cell-lines that transcriptomically match distinct tumor subtypes in a clinical cohort. Representative cell-lines were co-cultured to create “mixed-cell” models depicting inter-patient heterogeneity in CRPC, which were employed to assess drug combinations. When drug combinations previously tested in CRPC clinical cohorts were assessed to establish proof of concept, in vitro responses measured in our models concurred with their known clinical efficacy. Additionally, novel drug combinations computationally predicted to be efficacious in heterogeneous tumors were evaluated. They demonstrated preclinical efficacy in the mixed-cell models, suggesting they will likely benefit heterogeneous patient cohorts. Furthermore, we showed that the current practice of screening cell-lines/xenografts separately and aggregating their responses, failed to detect their efficacy. We believe that the application of our models will enhance the accuracy of preclinical drug assessment, thereby improving the success rate of subsequent clinical trials.