Inference of Cancer Drug Cross Resistance Using Only Single-Drug Exposure Data

Abstract

Combining drugs may prevent resistance in cancer treatment. However, the evolution of resistance to more than one drug simultaneously - cross resistance - remains a major obstacle to durable clinical response. Here, we show that we can identify the presence and strength of cross resistance using lineage tracing data from cell populations exposed only to each drug independently, removing the need for complex combinatorial treatment experiments. Simulations where the ground truth was known showed that the method accurately recovered the true underlying resistance dynamics. We applied the framework to breast cancer, lung cancer and lymphoma datasets, quantifying cross resistance across diverse therapy types. Our predictions align with biological expectations and reveal that even drugs with shared putative targets can exhibit variable cross resistance across pre-clinical models. Finally, we used inferred cross resistance strengths to evaluate treatment strategies, predicting that high cross resistance contributes to the limited efficacy of switching between CDK4/6 inhibitors. By enabling the identification of cross resistance without requiring multi-drug exposure experiments, our framework provides a method for prioritising drug combinations and screening for novel resistance-minimising drug targets.