Nanomotion-Based Drug Sensitivity Prediction in Ovarian and Colon Cancer Cell Lines Using Machine Learning

Cancer drug resistance remains a critical challenge in oncology, demanding rapid and reliable diagnostic tools to assess tumor cell susceptibility to treatment. This study presents a nanomotion-based drug susceptibility testing (DST) approach, integrating nanoscale movement analysis with supervised machine learning to classify drug-sensitive and drug-resistant cancer cells. Using label-free, real-time nanomotion technology, we measured the dynamic responses of colon cancer (SW480) and ovarian cancer (A2780, A2780ADR) cells to doxorubicin under physiological conditions. Features extracted from nanomotion signals were used to train machine learning models, achieving 90.9% accuracy in distinguishing between doxorubicin-treated and untreated SW480 cells and 84.6% accuracy in classifying doxorubicin-sensitive and-resistant ovarian cancer cells. The model achieved perfect classification of resistant A2780ADR cells in an independent test set after only 4 h and 15 min of exposure to the drug. Unlike genetic tests that infer drug resistance from molecular markers or metabolic assays requiring extended incubation times, nanomotion-based DST provides a direct phenotypic readout, offering a faster, label-free alternative for assessing tumor cell responses. While further dataset expansion and model refinement are necessary to enhance generalizability, these results underscore the potential of nanomotion technology as a rapid, phenotypic DST for personalized oncology. By directly measuring the mechanical behavior of cancer cells in response to chemotherapy, this method could transform clinical decision-making, enabling faster, more precise treatment strategies to combat drug resistance in cancer.