Inference of lineage hierarchies, growth, and drug response mechanisms in cancer cell populations without tracking

Lineage hierarchies and plasticity regulate development and tissue homeostasis, while diverted lineage dynamics and aberrant phenotypic plasticity are among the causes of incomplete drug response and resistance in cancer. Knowing the dynamics of phenotypically heterogeneous populations is therefore central to understanding growth regulation principles and to rationally design therapeutic approaches anticipating drug-tolerant states. While lineage inference can be addressed by barcoding technologies, these approaches often yield average clonal behaviors that neglect the underlying phenotypic plasticity of individual cells. Directly observing single-ancestor pedigrees in multi-type populations remains an experimental challenge. To address these difficulties, we developed a method to infer active phenotypic transitions in a multi-type tumor or clone and to quantify them, solely relying on counting cell-type abundances. We demonstrate the effectiveness of our approach to address cancer phenotypic heterogeneity and drug tolerance in silico. We then perform experiments on cancer cell populations and infer growth mechanisms and transition probabilities.