Analysis of Gliding Molecular Robots using Machine Learning Image Detection
Ibuki KAWAMATA *1, Sidak Grewal SINGH1, Yiming GONG1, Chung Wing CHAN1, Rubaya Mst. RASHID1, Marie TANI1, Masatoshi ICHIKAWA1, 2, Akira KAKUGO1
1Kyoto University
2Hiroshima University
Understanding the dynamics of each element within a collective behavior is extremely important for advancing our understanding of active matter, where a large number of motile elements come together to emerge functions that cannot be achieved by individual elements alone. One example of such active matter system implemented at the molecular level is a system of microtubules that glide on a surface driven by kinesin molecular motors. Microtubule systems induce various emergent phenomena by controlling interactions between elements using DNA molecules. For example, we have successfully controlled the timing of microtubule movement, assembly, and disassembly using artificially designed DNA chemical reaction networks. While movies observed using fluorescence microscopy are used to analyze such experimental systems, conventional methods have limitations in terms of reproducibility and scalability, as they involve manual data extraction, image processing requiring human intervention, and coarse-grained information processing based on statistical assumptions. Therefore, we attempted to quantitatively evaluate the physical characteristics of the microtubule gliding system from individual-scale movements by performing automatic feature extraction using image recognition and tracking technology based on machine learning. In this presentation, we will discuss the comparison of microtubule characteristics before and after assembling, as well as the analysis results of the interaction between kinesin-fixed surface and microtubules. This research contributes to a continuous understanding of individuals and collections, which range across different physical scales, and is expected to contribute to future applications such as smart drug delivery and biosensors using active biomolecules.