Structure-informed dissection of cancer-specific intracellular and paracrine networks

了解癌细胞自主行为和向肿瘤招募促恶性亚群 微环境（TME）关键取决于准确且全面的细胞和 细胞间网络。项目 1 的目标是开发一种新颖的、集成的、经过广泛验证的 模型、操纵和剖析肿瘤微环境中细胞信号传导的框架，涉及 细胞外配体-受体相互作用与细胞内信号网络耦合。项目 1 将建立在 上一个 CSBC 资助期间产生的方法和结果，以应对多重挑战 (a) 利用新颖的方法扩展蛋白质-蛋白质相互作用 (PPI) 网络的结构信息预测 深度学习方法，(b) 基于时间依赖性分析改进信号转导网络 药物扰动测定，以及（c）阐明配体/受体介导的旁分泌相互作用网络 介导健康细胞向 TME 的招募（并可能重新编程），以产生促恶性细胞 环境。为了实现这些目标，重点将放在两种高度侵袭性的肿瘤上——结肠癌 腺癌 (COAD) 和胰腺导管腺癌 (PDAC)——其数据、模型、试剂、 分析工具是在上一个融资周期内产生的。 项目 1 基于三个具体目标。通过将深度学习方法整合到蛋白质-蛋白质 相互作用以及为癌症标志创建基于结构的网络，目标 1 将提供 3D- 整个项目 1 中拟议工作的结构背景。目标 2 将定义基于磷酸蛋白质组学的 细胞内信号网络并描述它们对药物扰动的反应。目标 3 将定义旁分泌—— 基于细胞间信号网络，并使用新型器官芯片平台对其进行验证。 项目 1 的影响将主要源自其创新方法，其中包括使用结构- 基于分析来模拟蛋白质相互作用网络；基于结构的建模与深度建模的集成 学习算法，包括蛋白质语言模型，为基本上所有相互作用提供模型 将在提案中进行预测和观察；基于磷酸蛋白质组学的磷蛋白的推论 提供细胞信号传导关键的时间依赖性和扰动敏感成分的活性；的 并入旁分泌信号；以及新颖的实验验证技术，包括匹配 磷酸化蛋白质组学和转录谱，以及肿瘤和正常细胞的生物工程 互连的微室可以更好地重现体内组织生理学。 项目 1 的主要交付成果是一个可质疑的整体模型，用于耦合细胞内和细胞间 通过对机制进行剖析，信号将成为整个中心的基础 有助于肿瘤相关细胞状态的稳定性，它们的配体/受体介导的与其他细胞的相互作用 TME 中的亚群及其药理学上可行的分子依赖性。

Understanding cancer cell-autonomous behavior and recruitment of pro-malignant subpopulations to the tumor microenvironment (TME) is critically dependent on the generation of accurate and comprehensive cellular and intercellular networks. The goal of Project 1 is to develop a novel, integrated, and extensively validated framework to model, manipulate, and dissect cell-cell signaling in the tumor microenvironment involving extracellular ligand-receptor interactions coupled to intracellular signaling networks. Project 1 will build on the methodologies and results generated during the previous CSBC funding period to address multiple challenges by (a) expanding structure-informed prediction of protein-protein interactions (PPI) networks by leveraging novel deep learning approaches, (b) improving signal transduction networks based on the analysis of time-dependent drug perturbation assays, and (c) elucidating ligand/receptor-mediated paracrine interaction networks that mediate recruitment—and possibly reprogramming—of healthy cells to the TME to create a pro-malignant environment. To accomplish these goals, the focus will be on two highly aggressive tumors—colon adenocarcinoma (COAD) and pancreatic ductal adenocarcinoma (PDAC)—for which data, models, reagents, and analytical tools were generated during the prior funding cycle. Project 1 is based on three specific aims. Through the integration of deep learning approaches to protein-protein interactions and the creation of structure-based networks for the Hallmarks of Cancer, Aim 1 will provide a 3D- structural context for the proposed work throughout Project 1. Aim 2 will define phosphoproteomics-based intracellular signaling networks and describe their response to drug perturbations. Aim 3 will define paracrine- based cell-cell signaling networks and validate them with a novel organs-on-a-chip platform. The impact of Project 1 will derive largely from its innovative approaches, which include the use of structure- based analyses to model protein interaction networks; the integration of structure-based modeling with deep learning algorithms, including Protein Language Models, to provide models for essentially all interactions that will be predicted and observed in the proposal; the inference of phosphoproteomics-based phosphoprotein activity to provide critical time-dependent and perturbation-sensitive components of cellular signaling; the incorporation of paracrine signaling; and novel experimental validation technologies including matched phosphoproteomic and transcriptional profiles, and the bioengineering of tumors and normal cells within interconnected micro-chambers to better recapitulate tissue physiology in vivo. The major deliverable for Project 1 is an interrogable and holistic model for coupled intra- and inter-cellular signaling which will serve as the foundation for the entire center by enabling the dissection of the mechanisms contributing to the stability of tumor-related cell states, their ligand/receptor-mediated interaction with other subpopulations in the TME, and their pharmacologically actionable molecular dependencies.