Modeling and Evolution of Biological Networks

该奖项支持生物网络建模和进化方面的理论研究和教育。进行的研究使用理论模型来解决遗传模式的进化问题，这些模型代表了遗传密码和可能的变化，同时选择具有“适应度函数”的有利变化。理论方法借鉴了计算机科学中的学习模型和统计物理学中的优化。这种理论机器通过从突变和选择提供的随机样本中学习生物网络的速度来预测生物网络。该方法假设生物中的网络可以逐步建立，并通过逐步增加适应度而不断增长。它们不一定是全局最优的。网络进化将用发育途径的中心分子组分的简化微分方程来建模，例如，转录因子、配体-受体相互作用、蛋白质-蛋白质复合物、激酶和磷酸化蛋白质等。进化需要适应度函数，并且快速进化由平滑的单调函数而不是锯齿状的景观促进。这与进化所有动物而不是特定门共有的模式网络的目标是雅阁的。一个合理的适应度函数量化了胚胎位置与基因表达模式的关系。本研究遵循这些想法的初步应用，在动物的周期分割（体节发生）的最早版本的方法发现了令人鼓舞的成功程度。所进行的努力具有更广泛的影响与科学和教育的后果。这项研究从凝聚态物理学的角度采用了这些方法，在洛克菲勒大学的一个专注于生物研究的环境中进行。参与研究的研究生获得了独特的跨学科教育，理论物理学的基础沉浸在生物科学的研究环境中。这项研究对科学界的贡献超出了出版和通常的传播形式。建模和分析中使用的生物信息学工具被打包到网站中，以供广泛传播。 在一篇论文中，该团队合作通过全基因组重测序追踪人类患者体内耐药金黄色葡萄球菌（"超级细菌"）的进化，说明了所开发的工具与真实的世界医学的相关性。非技术性总结：该奖项支持生物网络建模和进化方面的理论研究和教育。该研究使用理论模型模拟遗传模式的进化，这些模型代表遗传密码和可能的变化，同时选择具有“适应度函数”的有利变化。理论方法借鉴了计算机科学中的学习模型和统计物理学中的优化。 这种理论机器根据从突变和选择提供的随机示例中学习生物网络的速度来预测生物网络。该方法假设生物中部署的网络可以逐步建立，并通过逐步增加适应度而不断增长。它们不一定是全局最优的。 网络进化将模拟与简化方程的发展途径的中心分子组成部分。本研究遵循这些想法的初步应用，在动物的周期分割（体节发生）的最早版本的方法发现了令人鼓舞的成功程度。所进行的努力具有更广泛的影响与科学和教育的后果。这项研究从凝聚态物理学的角度采用了这些方法，在洛克菲勒大学的一个专注于生物研究的环境中进行。参与研究的研究生获得了独特的跨学科教育，理论物理学的基础沉浸在生物科学的研究环境中。 这项研究对科学界的贡献超出了出版和通常的传播形式。建模和分析中使用的生物信息学工具被打包到网站中，以供广泛传播。 在一篇论文中说明了开发的工具与真实的世界医学的相关性，该小组合作通过全基因组重测序跟踪人类患者中耐药金黄色葡萄球菌（"超级细菌"）的进化。

TECHNICAL SUMMARY:This award supports theoretical research and education in modeling and evolution of biological networks. The research undertaken addresses evolution of genetic patterns using theoretical models that represent the genetic code and the changes that are possible while selecting changes as favorable with a ''fitness function.'' The theoretical approach draws on analogies with learning models in Computer Science and optimization in Statistical Physics. This theoretical machinery predicts biological networks by how rapidly they can be learned from the random examples provided by mutation and selection. The approach assumes the networks in living things can be built incrementally and grow continually by stepwise increases in fitness. They are not necessarily global optimum. Network evolution will be modeled with simplified differential equations for the central molecular components of developmental pathways, e.g., transcription factors, ligand-receptors interactions, protein-protein complexes, kinases and phosphorylated proteins etc. Evolution requires a fitness function and rapid evolution is facilitated by a smooth monotone function, not a jagged landscape. This is in accord with the goal of evolving patterning networks common to all animals, not specific phyla. A plausible fitness function quantifies how well embryonic position is related to gene expression patterns. This research follows a preliminary application of these ideas to periodic segmentation in animals (somitogenesis) where the earliest versions of the approach found an encouraging degree of success.The effort undertaken has broader impacts with both scientific and educational consequences. The research, employing the approaches from a condensed matter physics perspective, takes place at Rockefeller University in an environment focused on biological studies. Graduate students involved in the research gain a unique interdisciplinary education with the foundations of theoretical physics immersed in the research environment of biological sciences. The research makes contributions to the scientific community beyond publishing and the usual forms of dissemination. The bioinformatics tools employed in the modeling and analysis are packaged into web sites for broad dissemination. The relevance of the tools developed to real world medicine were illustrated in a paper where this group collaborated in tracking the evolution of drug resistant Staphylococcus aureus (''Super bugs'') in a human patient by whole genome resequencing.NONTECHNICAL SUMMARY:This award supports theoretical research and education in modeling and evolution of biological networks. The research models evolution of genetic patterns using theoretical models that represent the genetic code and the changes that are possible while selecting changes as favorable with a ''fitness function.'' The theoretical approach draws on analogies with learning models in Computer Science and optimization in Statistical Physics. This theoretical machinery predicts biological networks based on how rapidly they can be learned from the random examples provided by mutation and selection. The approach assumes the networks deployed in living things can be built incrementally and grow continually by stepwise increases in fitness. They are not necessarily global optimum. Network evolution will be modeled with simplified equations for the central molecular components of developmental pathways. This research follows a preliminary application of these ideas to periodic segmentation in animals (somitogenesis) where the earliest versions of the approach found an encouraging degree of success.The effort undertaken has broader impacts with both scientific and educational consequences. The research, employing the approaches from a condensed matter physics perspective, takes place at Rockefeller University in an environment focused on biological studies. Graduate students involved in the research gain a unique interdisciplinary education with the foundations of theoretical physics immersed in the research environment of biological sciences. The research makes contributions to the scientific community beyond publishing and the usual forms of dissemination. The bioinformatics tools employed in the modeling and analysis are packaged into web sites for broad dissemination. The relevance of the tools developed to real world medicine were illustrated in a paper where this group collaborated in tracking the evolution of drug resistant Staphylococcus aureus (''Super bugs'') in a human patient by whole genome resequencing.