Revealing mechanisms of specificity and adaptability in molecular information processing through data-driven models

项目摘要/摘要 地球上生命的成功源于它使用分子来携带信息和 实现控制化学的算法，允许有机体对其 环境。传递信息和自适应响应的能力最终取决于 能够从众多分子信号中选择性地识别一个分子信号的分子系统 其他类似的信号。信号可以是一个分子(分子特异性)， 分子(组合专一性)或时变浓度模式(时间 专一性)。此外，这些分子系统需要保持适应性，以切换其 根据需要的专一性。这项提议的中心目标是理解分子 通过建立分子的预测模型为信息处理奠定基础， 组合和时间的专一性以及这种专一性的适应性。我们会 结合生物物理基础模型、信息论和动力系统框架 用于信号传递以创建分子、组合和时间特异性的数据驱动模型。 我们将在三个尺度上追问：(1)分子特异性：蛋白质如何像 抗体识别特定的伴侣，如病毒尖峰蛋白上的表位，但仍可以 通过突变迅速改变其特异性？我们将开发一种生物物理上知情的 基于机器学习的工具箱，用于利用Directed观察到的进化轨迹 进化实验以了解这种适应性的起源。(2)组合 特异性：BMP和转化生长因子-β等发育途径如何决定特定的配体 决定细胞命运的组合，即使每个配体混杂地结合多个 受体？我们将使用分子协作性的信息论框架来建立 多对多信令架构模型，并利用细胞图谱数据和实验进行验证 共同表达受体亚基的新组合。(3)时间专用性：如何 分子电路对特定的浓度随时间变化的模式做出反应，但不对其他浓度模式做出反应 细胞因子信号和昼夜节律？我们将开发动力系统-理论指导 随机共振模型，允许核因子-kB对否则无法检测的水平做出反应 细胞因子和昼夜节律-代谢耦合模型以了解细胞如何缓冲 营养波动。我们的工作的特点是结合了生物物理模型，提供了 对统计模型的理解和洞察，能够更好地利用现代高 吞吐量数据，并提供预测能力。此外，我们的推理工具箱和 相关的理论-实验工作流程可以被其他实验室用于类似的概念 关于替代系统的问题，如抗体和刺激性的分子特异性 蛋白质、转化生长因子-β途径的组合特异性或表皮生长因子的时间特异性 分别为上面的三次冲刺发出信号。

Project summary/abstract The success of life on earth derives from its use of molecules to carry information and implement algorithms that control chemistry, allowing organisms to respond adaptively to their environment. The ability to transduce information and respond adaptively ultimately relies on molecular systems being able to selectively recognize one molecular signal from among many other similar signals. The signal could be a molecule (molecular specificity), a combination of molecules (combinatorial specificity), or a time varying concentration pattern (temporal specificity). Further, these molecular systems need to remain adaptable to switch their specificity as needed. The central goal of this proposal is to understand the molecular basis of information processing by building predictive models of molecular, combinatorial and temporal specificity and adaptability of such specificity. We will combine biophysically grounded models, information theory and dynamical systems frameworks for signaling to create data-driven models of molecular, combinatorial and temporal specificity. We will pursue questions on three scales: (1) molecular specificity: how do proteins like antibodies recognize a specific partner, such as an epitope on a viral spike protein, and yet can rapidly change its specificity through mutations? We will develop a biophysically informed machine learning-based toolbox to exploit evolutionary trajectories observed in directed evolution experiments to understand the origin of such adaptability. (2) combinatorial specificity: how do developmental pathways like BMP and TGF-beta resolve specific ligand combinations to determine cell fate, even though each ligand promiscuously binds multiple receptors? We will use an information theory framework for molecular cooperativity to build models of many-many signaling architectures and validate using cell atlas data and experiments that co-express novel combinations of receptor subunits. (3) temporal specificity: how do molecular circuits respond to specific time-varying patterns of concentrations but not others in cytokine signaling and in circadian rhythms? We will develop dynamical systems-theory guided models of stochastic resonance that allow NF-kB to respond to otherwise undetectable levels of cytokines and models of circadian clock-metabolism coupling to understand how cells buffer nutrient fluctuations. Our work is distinguished by combining biophysical models which provide understanding and insight with statistical models that are better able to leverage modern high- throughput data and provide predictive power. In addition, our inference toolboxes and related theory-experiment workflows can used by other labs for similar conceptual questions about alternate systems, such as, molecular specificity for antibodies and spike proteins, combinatorial specificity in the TGF-beta pathway or temporal specificity in EGF signaling respectively for the three thrusts above.