Automated particle tracking and stochastic modeling of molecular motion in submicron living systems

本提案概述了我们建立“发现管道”的计划，包括数据，建模和预测，以研究粘膜免疫学和细胞生物学中的问题。特别是，我们研究“粒子”（即分子，病原体，细胞等）在微米尺度，细胞内和细胞外环境中的随机运动。这条管道将提供关于拥挤、粘度和弹性以及隐藏的分子动力学如何影响生物过程等基本问题的答案。虽然这项工作的重点是使用数学建模来获得对生物系统的机械洞察，但它也将包括统计学、机器学习和计算机科学工具的开发。颗粒的运动有时是研究的直接焦点（例如，沙门氏菌在粘液屏障中的运动）。然而，我们也使用粒子作为分子“探针”来推断粒子局部环境的隐藏特性，如粘度、流动、拥挤、弹性和约束。在任何一种情况下，随机建模都可以机械地将粒子运动从数据（运动的视频记录）与跨越时间和空间尺度的预测以及隐藏属性的推断联系起来。我们将首先建立发现管道的基础：一个粒子跟踪器，它使用神经网络进行自动化，并使用贝叶斯网络连接到模型。该方法的测试和验证将与涉及现有实验数据集的多个项目协调进行。结果是(i)量化和阐明特定生物学问题的答案，以及（ii）开发和验证我们的发现管道。长期目标是使世界各地的研究团队都可以使用并采用该管道。

This proposal outlines our plan to build a 'discovery pipeline'-encompassing data, modeling, and prediction-to study problems in mucosal immunology and cell biology. In particular, we study stochastic motion of 'particles' (i.e., molecules, pathogens, cells, etc) in micron-scale, and in intracellular and extracellular environments. This pipeline will produce answers to fundamental questions about how crowding, viscosity and elasticity, and hidden molecular kinetics impact biological processes. While the focus of the work is using mathematical modeling to gain mechanistic insight into biological systems, it will also include development of tools from statistics, machine learning, and computer science. The motion of particles is sometimes the direct focus of the research (e.g., Salmonella motion in mucus barriers). However, we also use particles as molecular 'probes' to infer hidden properties of the particle's local environment, such as viscosity, flow, crowding, elasticity, and confinement. In either case, stochastic modeling can mechanistically link particle motion from data (video recordings of motion) to predictions across time and space scales and for inference of hidden properties. We will first build the foundation of the discovery pipeline: a particle tracker that uses a neural network for automation and Bayesian networks to connect to models. Testing and validation of the approach will be conducted in coordination with multiple projects involving existing experimental data sets. The outcomes are (i) quantifying and elucidating answers to specific biological questions, and (ii) developing and validating our discovery pipeline. The long-term goal is to make the pipeline accessible to and adopted by research teams world-wide.