CAREER: Learning stochastic spatiotemporal dynamics in single-molecule genetics

测量哪些基因在细胞中表达的能力彻底改变了我们对生物系统的理解。发现范围从确定是什么使不同的细胞类型独特（例如，皮肤细胞与脑细胞）到疾病如何从基因突变中产生。这种基因表达数据现在是每个细胞生物学家工具箱中普遍使用的工具。然而，从这些数据中可靠地提取洞察力的数学理论已经落后于收获这些数据的技术的惊人进步。本项目将为基因表达成像数据分析提供关键的理论基础。这些进步从理论到实践，包括开发数学模型和机器学习方法，这些方法将与实验合作者的数据一起使用。总而言之，该项目旨在为研究基因表达的空间成像数据创造新的技术标准，并使新的生物学和生物医学见解得以揭示。此外，这项拟议的研究将包括跨学科的研究生和当地社区大学的本科生，以培养数据科学、生物学和数学不断发展的交叉领域的下一代科学家。除了研究活动，该项目还将创建导师网络，以支持第一代学生科学家追求STEM劳动力多样化。支持的研究是通过随机反应-扩散模型研究单分子基因表达空间模式的综合项目。关键目标是将这些模型和观测结果之间的数学联系概括为空间点过程。新的理论将纳入必要的因素来描述空间基因表达在亚细胞和多细胞尺度，包括各种反应，空间运动，和几何效应。该项目还将建立仅从单个粒子位置的快照推断随机速率的反问题的推理统计理论。对参数可辨识性、最佳实验设计和模型选择的研究将确保稳健和可靠的推断。作为已开发理论的补充，该项目将实施和基准测试有效执行大规模统计推断的尖端方法，包括变分贝叶斯蒙特卡罗和物理信息神经网络。这项工作的最终成果将被打包到开源软件中，该软件可以从多基因组织规模的数据集中推断出可解释的生物物理参数。该职业奖由数学科学部的数学生物学和统计项目以及生物技术理事会分子和细胞生物科学部的细胞动力学和功能集群共同资助。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The ability to measure which genes are expressed in cells has revolutionized our understanding of biological systems. Discoveries range from pinpointing what makes different cell types unique (e.g., a skin vs. brain cell) to how diseases emerge from genetic mutations. This gene expression data is now a ubiquitously used tool in every cell biologist’s toolbox. However, the mathematical theories for reliably extracting insight from this data have lagged behind the amazing progress of the techniques for harvesting it. This CAREER project will develop key theoretical foundations for analyzing imaging data of gene expression. The advances span theory to practice, including developing mathematical models and machine-learning approaches that will be used with data from experimental collaborators. Altogether, the project aims to create a new gold standard of techniques in studying spatial imaging data of gene expression and enable revelation of new biological and biomedical insights. In addition, this proposed research will incorporate interdisciplinary graduate students and local community college undergraduates to train the next generation of scientists in the ever-evolving intersection of data science, biology, and mathematics. Alongside research activities, the project will create mentorship networks for supporting first-generation student scientists in pursuit of diversifying the STEM workforce. The supported research is a comprehensive program for studying single-molecule gene expression spatial patterns through the lens of stochastic reaction-diffusion models. The key aim is to generalize mathematical connections between these models and their observation as spatial point processes. The new theory will incorporate factors necessary to describe spatial gene expression at subcellular and multicellular scales including various reactions, spatial movements, and geometric effects. This project will also establish the statistical theory of inference on the resulting inverse problem of inferring stochastic rates from only snapshots of individual particle positions. Investigations into parameter identifiability, optimal experimental design, and model selection will ensure robust and reliable inference. In complement to the developed theory, this project will implement and benchmark cutting-edge approaches for efficiently performing large-scale statistical inference, including variational Bayesian Monte Carlo and physics-informed neural networks. The culmination of this work will be packaged into open-source software that infers interpretable biophysical parameters from multi-gene tissue-scale datasets.This CAREER Award is co-funded by the Mathematical Biology and Statistics Programs at the Division of Mathematical Sciences and the Cellular Dynamics & Function Cluster in the Division of Molecular & Cellular Biosciences, BIO Directorate.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.