Skip to main content
Fig. 5 | Journal of Nanobiotechnology

Fig. 5

From: Growth rate-dependent flexural rigidity of microtubules influences pattern formation in collective motion

Fig. 5

Distinct patterns formed by MTs with different flexural rigidities. A MT gliding assays conducted within a reconstructed MT-kinesin system in the presence of methylcellulose. B A deep learning CNN model, including global average pooling (GAP) layer, final connected (FC) layer, and MT class output, is constructed to classify the two MTs: softer-MTs and stiffer-MTs. The two groups of MTs are polymerized at different tubulin concentration of 30 μM and 100 μM, which are named as softer-MTs (κ = 0.27 × 10−23 N m2) and stiffer-MTs (κ = 0.80 × 10−23 N m2), respectively. C From the identical isotropic state, softer-MTs and stiffer-MTs form distinctive patterns at the nematic phase. The addition time of ATP was set as 0 min. Scale bar = 50 μm. D Confusion matrix based on the trained CNN classifier. Here, 200 MT pattern images are categorized using the classifier. E The classification strategy of CNN classifier is visually explained using Score-CAM

Back to article page