CCF: AF: Small: Novel Stochastic Optimization Algorithms to Advance the Treatment of Dynamic Molecular Systems

该项目旨在推进计算生物学中的算法研究。重点是需要分子表征的问题，如对细胞内部工作至关重要的蛋白质和肽。这些分子经常在不同的结构构型之间切换，以与细胞中的不同分子伴侣相容并相互作用。计算蛋白质或肽在生物活性中的不同构型对于理解这些分子在健康和患病细胞中如何运作至关重要。然而，这样做，提出了许多计算的挑战，在这个项目中的研究旨在提高我们的能力，以满足这些challenges.The项目概括的问题，获得可能不同的功能状态的综合视图，获得一个不同的合奏约束满足结构。提出了新的随机优化技术来处理这些分子中产生的高维变量空间，以及与这些空间相关的能量表面的非线性和多峰性。这些技术包括混合和结构化，以有效地平衡勘探和开发之间的计算资源。重点是在并行架构中混合这两者。新的多目标分析解决了能量函数中的固有噪声。积累的生物物理知识驱动新的变异算子的设计，有效的全球和局部搜索。该研究以软件的形式提供，以促进随机优化的发展，同时允许计算生物学家使用尖端技术对感兴趣的特定系统进行建模。该研究支持跨计算机科学，生物工程和神经科学的协调教学和推广活动，整合博士后，大学和大学预科生。

This project aims to advance algorithmic research in computational biology. The focus is on problems that demand characterizations of molecules, such as proteins and peptides that are central to the inner workings of cell. These molecules often switch between different structural configurations to be compatible and interact with different molecular partners in the cell. Computing the different configurations a protein or peptide takes on for biological activity is central to understanding how these molecules operate in the healthy and diseased cell. Doing so, however, poses many computational challenges, and the research in this project aims to advance our ability to meet these challenges.The project generalizes the problem of obtaining a comprehensive view of possibly diverse functional states as that of obtaining a diverse ensemble of constraint-satisfying structures. Novel stochastic optimization techniques are put forth to deal with the high-dimensional variable spaces arising in these molecules, as well as the nonlinearity and multimodality of the energy surface associated with these spaces. The techniques include hybridization and structurization to effectively balance computational resources between exploration and exploitation. The focus is on blending the two in parallel architectures. Novel multi-objective analysis addresses the inherent noise in energy functions. Accumulated biophysical knowledge drives the design of novel variation operators for effective global and local search. The research is made available as software to facilitate advancement of stochastic optimization while allowing computational biologists to model specific systems of interest with cutting-edge techniques. The research supports a number of coordinated teaching and outreach activities across computer science, bioengineering, and neuroscience, integrating postdoctoral, college, and pre-college students.