Data Driven Analysis Of The Cell Dynamics Based On The Microscopy Imaging

The analysis of the temporal dynamics of single cells is important for the understanding of the biological self-organized regulations and functions.On one hand,the cell heterogeneity as well as the cell-cell interactions make it essential to carry out the researches in single cell level,one the other hand,time-lapse tracking is also significant for analyzing the state transition dynamics in biological progressions.Given the high throughput,high spatiotemporal resolution and the nondestructive features,the microscopic imaging analysis has become one of the most important methods for in-situ time-series single cell analysis.However,three-dimension(cell-to-cell varieties,time and space)detections lead to large data collections,which raise higher requirements on the standardization,automation and integration of the data analysis procedure.Furthermore,given the complexity of the temporal dynamics of the single cells,hypothesis-centered researches based on subjective modeling or assumptions may bring bias to the downstream analysis.This paper tries to address these issue with a data-driven perspective.Instead of predefining the models or analysis targets of the system,we tried to describe the system in an unbiased,objective and quantitative manner by collecting big data and introducing data visualization as well as data mining methods from computer science.More specifically,the paper contains the following researches:1.An unsupervised machine learning-based single particle trajectory analysis method was developed to reveal the spatiotemporal heterogeneity and state transition of the single particles in cell environment.Taking nanoparticle-cell interactions as an example,the method achieved transmembrane site identification as well as the hopdiffusion behavior analysis of the nanoparticles without any pretraining.We also proved the generality of the methods by applying it to the cell migration trajectory analysis.This method can aid the researches of the transmembrane dynamics and mechanism of nanocargoes.2.A collective behavior analysis method was developed to uncover the collective regulation patterns in system level.Taking the lysosomes in single cell as an example,the method tracked the motion of individual lysosomes and quantitatively characterized the collective spatiotemporal dynamics of the lysosomes.The spatial heterogeneity and organizations were revealed and we also investigated the relationship between the spatiotemporal collective dynamics of the lysosomes and the cell states.The method provides a new perspective for characterizing the cell activities.3.A cell morphology tracking method was developed to reveal the cell state transition dynamics in cell mitosis,differentiation and migration.The method tracked the mitosis process of He La cells and the polarization process of the bone marrow derivate macrophages(BMDMs).It identified the state transition point by deep learning and cell morphology trajectory construction.Combing the spatial and morphological information,we quantified the migration activities of the BMDMs before and after the M1 polarization.