Cotner, K. L. | Search bioRxiv

bioRxiv2025

MechanoAge, a machine learning platform to identify individuals susceptible to breast cancer based on mechanical properties of single cells

BackgroundExisting breast cancer risk models inadequately identify individuals at latent risk, particularly among women without known genetic mutations or family history. Risk is often underestimated or overestimated due to reliance on population-level data and neglect of cellular aging and mechanobiological alterations. MethodsWe profiled primary human mammary epithelial cells (HMECs) from women of varying ages and risk backgrounds using mechano-node-pore sensing (mechano-NPS), a high-throughput microfluidic platform that captures single-cell mechanical properties. Using machine learning, we developed a classifier, MechanoAge, to predict age-related mechanical phenotypes and introduce a novel index, mechano-RISQ, to quantify deviations linked to breast cancer risk. We further assessed the cytoskeletal protein keratin 14 (KRT14) as a molecular mediator of these mechanical states through overexpression and knockdown experiments. FindingsCells from younger women carrying BRCA1/2 mutations or with a family history of breast cancer exhibited accelerated mechanical aging compared to age-matched controls. Elevated mechano-RISQ scores reflected an increased proportion of cells with "older" mechanical profiles. KRT14 overexpression induced an aged mechanical phenotype in younger cells, while knockdown partially reversed this state in older cells. CyTOF profiling and modeling showed KRT14 modulation impacted protein expression signatures associated with aging and risk, particularly in luminal cells. InterpretationMechanical properties of breast epithelial cells reflect biologic aging and cancer susceptibility. Mechano-RISQ offers a new approach for identifying individuals at elevated risk, especially among average-risk populations, and may complement existing risk models by incorporating biophysical measures of epithelial aging.

cancer biology↗

bioRxiv2021

Evaluating sources of technical variability in the mechano-node-pore sensing pipeline and their effect on the reproducibility of single-cell mechanical phenotyping

Cellular mechanical properties can reveal physiologically relevant characteristics in many cell types, and several groups have developed microfluidics-based platforms to perform single-cell mechanical testing with high throughput. However, prior work has performed only limited characterization of these platforms technical variability and reproducibility. Here, we evaluate the repeatability performance of mechano-node-pore sensing, which is a single-cell mechanical phenotyping platform developed by our research group. We measured the degree to which device-to-device variability and semi-manual data processing affected this platforms measurements of single-cell mechanical properties, and we demonstrated high repeatability across the entire technology pipeline even for novice users. We then compared results from identical mechano-node-pore sensing experiments performed by researchers in two different labs with different analytical instruments, demonstrating that the mechanical testing results from these two locations are in agreement. Our findings quantify the expectation of technical variability in mechano-node-pore sensing even in minimally experienced hands. Most importantly, we find that the repeatability performance we measured is fully sufficient for interpreting biologically relevant single-cell mechanical measurements with high confidence.

bioengineering↗