Nanoscale stress microscopic analysis by fusion of machine learning and atomic force microscope spectroscopy

通过机器学习和原子力显微镜光谱融合进行纳米级应力微观分析

• 批准号: 19K15398
• 负责人: 王 洪欣
• 金额: $ 2.16万
• 依托单位: National Institute for Materials Science
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Early-Career Scientists
• 财政年份: 2019
• 资助国家: 日本
• 起止时间: 2019-04-01 至 2020-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-19K15398/
• 关键词: stress; machine learning; spectroscopy; ナノスケール応力; 機械学習; 原子間力顕微鏡スペクトロスコピー

Cells rely on cytoskeletal prestress to sense and transmit external force signals to induce chemical and mechanical responses. The intracellular spatial distribution of prestress thus determines the basic cytoskeletal functionalities of maintaining cellular homeostasis, which is lacking in diseased cell status including cancer. In this work, we measured spatially resolved prestress produced by local actomyosin machinery in a living cell with multivariable AFM force spectroscopy. The mechanical stimuli applied by the AFM probe were found to induce cellular prestress responses, in both normal cells and cancer cells, but in distinctively different manner at time scales from sub-second to hours. We demonstrated that a simple machine learning algorithm can be applied on un-segmented prestress distribution data to differentiate cancer cells and normal cells with accuracy above 95%. It promises a new biomarker for cancer cytology diagnosis on difficult cell specimens with high morphological similarities.

细胞依靠细胞骨架预应力来感知和传递外力信号，从而诱导化学和机械反应。因此，预应力的细胞内空间分布决定了维持细胞稳态的基本细胞骨架功能，这在包括癌症在内的病变细胞状态中是缺乏的。在这项工作中，我们测量了空间分辨预应力产生的局部肌动球蛋白机械在活细胞与多变量AFM力谱。由AFM探针施加的机械刺激被发现诱导细胞的预应力反应，在正常细胞和癌细胞，但在明显不同的方式在从亚秒到小时的时间尺度。我们证明了一种简单的机器学习算法可以应用于未分割的预应力分布数据，以区分癌细胞和正常细胞，准确率超过95%。它有望为具有高度形态相似性的困难细胞标本的癌症细胞学诊断提供新的生物标志物。

项目成果

High spatial and temporal resolution mapping of active stress on the surface of living material.

生物材料表面主动应力的高空间和时间分辨率映射。

• 发表时间: 2019
• 作者: S. Hatakeyama;S. Hiura;J. Takayama;and A. Murayama;Hongxin Wang
• 通讯作者: Hongxin Wang