AF: SMALL: Computational Framework for Characterizing Protein Conformational Landscapes

蛋白质几乎参与了生命的每一个过程。多年来，蛋白质结构、动力学和功能之间的关系一直是实验生物学家和计算机科学家面临的挑战，但由于蛋白质折叠、结合和结构域运动等事件的复杂性，许多研究问题仍未解决。由于蛋白质折叠和结合等事件的高复杂性和高维性，现有的分析蛋白质构象空间结构和功能特性的计算方法受到限制。教育和推广活动将通过以下方式实施：a)与马萨诸塞大学波士顿分校内外的研究人员进行跨学科合作。b)培训和指导本科生和研究生的研究，包括妇女和来自科学领域代表性不足群体的学生。将通过通往学士学位项目、UMB女性科学俱乐部和麻省大学波士顿分校IMSD项目的桥梁项目，积极开展与女性和弱势群体学生的合作。所提出的算法利用蛋白质的几何和生物物理特性来有效地表征它们的构象空间，并检测可能在功能上重要但难以通过实验确定的有趣区域。本文将探讨以下三个相关的研究问题：1。表征蛋白质的灵活性和约束：通过结合计算几何和生物物理学的概念，将开发有效采样蛋白质构象景观的方法。这一部分的重点将放在刚性分析和概率方法上。有效的低维构象空间表示：将测试有效的特征选择和降维技术，以可靠地表示构象空间。这些方法使用较少的变量表示高维、复杂的数据，同时保留基本信息。这将有助于蛋白质动力学的分析和表征。表征中间构象：将开发结合几何，数学和拓扑工具的强大算法方法，以有效地分析蛋白质构象空间并发现人口密集的区域。重点将放在聚类方法、异常值的检测以及处理噪声和多重约束。目标是确定蛋白质构象空间的重要结构和功能特性，如低能最小值、高能势垒和中间态的形状和数量。

Proteins are involved in virtually every process in life. The relationship between protein structure, dynamics and function has challenged experimental biologists and computer scientists for years, but many research questions remain open due to the complexity of events such as protein folding, binding and domain motion. Existing computational methods to analyze the structural and functional properties of protein conformational spaces are limited due to the high complexity and high dimensionality of events such as protein folding and binding. Educational and outreach activities will be implemented through the following: a) Interdisciplinary collaborations with researchers inside and outside UMass Boston.b) Training and mentoring the research of undergraduate and graduate students, including women and students from under-represented groups in science. Active work with women and students from under-represented groups will be pursued through the Bridges to the baccalaureate program, the UMB women in science club and the IMSD program at UMass Boston.Proposed algorithms exploit the geometric and biophysical properties of proteins to efficiently characterize their conformational space and detect interesting regions that may be functionally important but are hard to determine experimentally. The three following related research problems will be explored:1. Characterize Protein Flexibility and Constraints: Methods will be developed for effective sampling of protein conformational landscape by combining concepts from computational geometry and biophysics. The emphasis of this part will be on rigidity analysis and probabilistic methods.2. Effective Low-Dimensional Representation of the Conformational Space: Effective feature selection and dimensionality reduction techniques for reliable representation of the conformational space will be tested. These methods represent high dimensional, complex data using a smaller number of variables while preserving essential information. This will facilitate the analysis and characterization of protein dynamics.3. Characterizing Intermediate Conformations: Powerful algorithmic methods that combine geometry, mathematical and topological tool will be developed to effectively analyze the protein conformational spaces and discover highly populated regions. The emphasis will be on clustering methods, detection of outliers and handling noise and multiple constrains. The goal is to identify important structural and functional properties of protein conformational spaces such as the shape and number of low energy minima, high energy barriers and intermediate states.