Skip to main content
. 2021 Mar 15;11:5950. doi: 10.1038/s41598-021-85276-5

Figure 6.

Figure 6

(a) Non-invasive and invasive cancer cells showed different, yet heterogeneous, morphologies on non-aligned fibrous templates, and this difference was highlighted by (b) different YAP/TAZ states. (c) A neural network was constructed to classify cells as invasive or non-invasive based on morphology and/or YAP/TAZ levels. (d) Using morphology to classify cells resulted in 80.4% accuracy, while YAP/TAZ predicted cell classification with 90.2% accuracy. Using both morphology and mechanosensing increased accuracy to 95.1%, highlighting the utility of exploiting both morphology and mechanosensing in invasive cancer cells (**** p < 0.001, nMCF-7 = 177 and nMDA-MB-231 = 115 cells).