Intra-tumour heterogeneity (ITH) and cellular plasticity drive therapeutic resistance in melanoma. Although single-cell transcriptomics defines cell states, it is costly and slow. We hypothesized that simple biophysical parameters, cell surface (S) and volume (V), capture transcriptomic diversity and dynamic transitions. We developed ORIGAMI (mOnitoring heteRogeneity and plastIcity using cell Geometry scAling MetrIcs), a scalable flow cytometry framework to resolve melanoma cell states and guide rational combinations.

Human and murine melanoma cell lines, patient-derived xenografts (PDX), and syngeneic models were analyzed. Surface and volume were quantified using MemBrite and FSC-W, validated by confocal 3D reconstruction and Coulter measurements. Single-cell RNA-seq and proteomics on index-sorted cells correlated S/V metrics with molecular states. Drug-induced transitions were mapped as trajectories in S-V space. High-volume cells were targeted with lysosome-directed ionophores, and ORIGAMI-guided screening identified surface-reducing compounds. Therapeutic efficacy was evaluated in vitro and in vivo, including in NRAS-mutant PDX and immunocompetent models.

Transcriptomic states occupied discrete S-V regions: mesenchymal-like cells displayed high surface area, whereas therapy-induced persister cells exhibited increased volume. Principal transcriptomic components aligned with S and V, linking cell geometry to molecular programs. Targeted therapies induced reproducible trajectories in S-V space, uncovering vulnerabilities associated with surface expansion or lysosomal scaling in high-volume cells. Nigericin triggered ferroptosis in enlarged cells, while brefeldin A reduced surface area, promoted differentiation, and resensitized resistant populations. The quadruple combination delayed tumour growth, prolonged survival in NRAS-mutant PDX without overt toxicity, induced immunogenic cell death, and restored anti-PD-1 responsiveness. The S-V framework extended across cancer types, where elevated S/V ratios associated with increased metastatic potential and sensitivity to geometry-guided combinations.

Cell geometry metrics provide a simple, scalable surrogate for transcriptomic states. ORIGAMI captures cancer cell diversity and plasticity, predicts therapy-induced trajectories, and enables genotype-agnostic combinations. Modulating the S/V landscape offers a broadly applicable strategy to overcome adaptive resistance and enhance immunotherapy efficacy.

We thank collaborators not listed due to space limitations. We also thank Deepcell for their support and assistance.This work was supported by VIB and KU Leuven. A patent application covering ORIGAMI and geometry-guided combinations has been filed; some authors are inventors. Others declare no competing interests.