Network models of differentiation landscapes for angiogenesis and hematopoiesis

我的长期研究目标是开发数据驱动的数学模型来理解和控制 细胞分化。我的实验室的方法将单细胞转录数据集成到一个数学模型中， Hopfield模型，用于描述基因网络中的信号动力学，并模拟扰动对基因网络的影响 一组靶基因。Hopfield模型最初是作为大脑的数学模型开发的，它允许 将关联记忆模式直接映射到网络中的动态吸引子状态，从而系统 可以使用部分信息恢复大量内存。我们表示表型细胞状态的基本原理 所谓联想记忆是指当一个细胞通过表达一种新的基因模式进行分化时 在网络中，细胞依赖于一套由进化形成的内置联想记忆模式。与许多人不同 深度学习和其他机器学习方法，使用联想表示细胞决策过程 记忆提供可解释的信息，该信息可以整合先前存在的生物知识(例如，路径 信息)，以帮助阐明基本的生物规则。此外，这种方法超越了描述性的 单元格数据的分析。它为可以识别新的药物靶点的硅胶实验铺平了道路 为临床前和临床研究产生新的假设。 在接下来的五年里，我们计划应用我们的吸引子模型来帮助理解和控制两个 参与多种疾病的生物学过程：血管生成和造血。血管生成是 新血管的发育，是胚胎发育、成人血管动态平衡所必需的 组织修复。我们在血管生成方面的目标是识别器官特有的信号，这将提供新的机会 设计新的疗法来刺激或抑制病变器官(例如，患者的视网膜)的血管生成 患有老年性黄斑变性)，而不影响健康器官。在造血方面，我们将使用 我们的计算方法确定了从诱导多能干细胞中产生淋巴细胞的新方案 细胞(IPSC)。IPSC来源的淋巴细胞的生物制造对于许多应用非常重要，例如 有效生产用于CAR-T疗法的T细胞和开发癌症疫苗。 我们的电子实验确定的新靶标和靶标组合可能会导致新的 治疗许多疾病，包括癌症、失明、关节炎、牛皮癣和许多其他缺血性疾病， 炎症性、感染性和免疫性疾病。同样重要的是，这个Mira项目将使我们能够继续 与我们的合作者网络合作，与更广泛的生物医学领域共享创新的软件工具 社区，并进一步为培养一支具有强大的跨学科卫生科学队伍做出贡献 计算和数学技能。

My long-term research goal is to develop data-driven mathematical models to understand and control cell differentiation. My lab’s approach integrates single-cell transcriptomics data in a mathematical model, the Hopfield model, to describe the signaling dynamics in gene networks and simulate the effect of perturbations on sets of target genes. Originally developed as a mathematical model of the brain, the Hopfield model allows for a direct mapping of associative memory patterns into dynamical attractor states in a network, so that the system can recover a host of memories using partial information. Our rationale for representing phenotypic cell states as associative memories is that when a cell “decides” to differentiate by expressing a new pattern of genes in a network, the cell relies on a set of built-in associative memory patterns shaped by evolution. In contrast to many deep learning and other machine learning methods, representing cellular decision processes using associative memories provides interpretable information that can integrate pre-existing biological knowledge (e.g., pathway information) to help elucidate fundamental biological rules. Moreover, the approach goes beyond a descriptive analysis of single-cell data. It paves the way for in-silico experiments that could identify new drug targets and generate new hypotheses for pre-clinical and clinical investigation. In the next five years, we plan on applying our attractor models to help understand and control two inter- playing biological processes involved in many diseases: angiogenesis and hematopoiesis. Angiogenesis is the development of new blood vessels and is required for embryonic development, adult vascular homeostasis, and tissue repair. Our goal in angiogenesis is to identify organ-specific signals, which will provide new opportunities to design new therapeutics to stimulate or inhibit angiogenesis in diseased organs (e.g., retinas of patients suffering from age-related macular degeneration) without affecting healthy organs. In hematopoiesis, we will use our computational approaches to identify new schemes to generate lymphocytes from induced pluripotent stem cells (iPSC). The bio-manufacturing of iPSC-derived lymphocytes is important for many applications, such as effective production of T cells for CAR-T therapies and development of cancer vaccines. The new targets and target combinations identified by our in-silico experiments could lead to novel therapeutics for many diseases, including cancer, blindness, arthritis, psoriasis, and many other ischemic, inflammatory, infectious, and immune disorders. Just as important, this MIRA project will allow us to continue working with our network of collaborators, keep sharing innovative software tools with the broader biomedical community, and further contribute to the training of an interdisciplinary health sciences workforce with strong computational and mathematical skills.