Next Generation Forcefield for Internal Coordinate Mechanics

描述（由申请人提供）：使用内部坐标的分子建模，而不是更传统的笛卡尔方法，由于其显著的计算效率优势而获得认可。Molsoft的ICM（内部坐标力学）是为数不多的使用内部坐标从头开始构建的分子建模程序之一。ICM可能是目前唯一一个在蛋白质结构预测、肽与配体构象分析、蛋白质/蛋白质、蛋白质/配体对接、虚拟筛选等领域应用广泛的软件程序。在任何分子力学应用中，力场的准确性是成功建模的基础。在内部坐标中工作，特别是在扭转中，对模拟中使用的力场有特定的要求。事实上，当笛卡尔力场能量（错误地）用于评估在扭转空间中产生的构象时（在共价几何冻结的情况下），由于笛卡尔力场参数是在松弛的共价几何结构上优化的，因此能量最小值的相对能量和旋转障碍的较大误差是不可避免的。康奈尔大学Harold Scheraga教授的团队最近开发并发表的ECEPP05力场代表了扭转空间力场发展的重要一步。重要的是，专门衍生的与其他非键和键的溶剂化参数也被开发出来，并处于出版的后期阶段。该项目的第一阶段将重点关注ECEPP05在ICM平台中的实施和集成。在第一阶段结束时，我们期望有一个ICM的发布，允许用户在他们的蛋白质模拟中应用新的力场。我们将在环路建模中进一步测试新力场的性能。我们已经在ICM中实现了一个基于蒙特卡罗最小化的环路仿真协议。该方案在内部基准测试中显示出有希望的结果，我们期望使用新的力场在这类重要的多肽模拟中达到更高的精度。该提案的长期（第二阶段）目标是进一步将ECEPP05扩展为可应用于小分子，特别是对接的通用力场。其他扩展将包括DNA和金属蛋白建模参数。

DESCRIPTION (provided by applicant): Molecular modeling using internal coordinates, rather than more traditional Cartesian approaches, is gaining recognition because of its significant computational efficiency advantages. Molsoft's ICM (Internal Coordinate Mechanics) is among the very few molecular modeling programs built from the ground up using internal coordinates. ICM is probably the only such software program with the breadth of application fields including protein structure prediction, peptide and ligand conformational analysis, protein/protein and protein/ligand docking and virtual screening. In any molecular mechanics application, the accuracy of the force-field is the foundation for successful modeling. Working in internal coordinates, in particular torsions, imposes specific requirements on the force-fields used in the simulations. Indeed, when Cartesian force-field energy is (mis)used to evaluate conformations generated in torsion space (with the covalent geometry frozen), large errors in the relative energies of energy minima as well as rotation barriers are inevitable because the Cartesian force-field parameters are optimized on structures with relaxed covalent geometry. The ECEPP05 force-field recently developed and published by Prof. Harold Scheraga's group at Cornell University represents a major step forward in the development of torsion space force-fields. Importantly, solvation parameters specifically derived to work in conjunction with other non-bonded and bonded terms were also developed and are in the late stages of publication. Phase I of this project will focus on the implementation and integration of ECEPP05 into the ICM platform. At the end of Phase I, we expect to have a release of ICM that will allow users to apply the new force-field in their protein simulations. We will further test the performance of the new force-field in loop modeling. We already have a Monte-Carlo minimization-based loop simulation protocol implemented in ICM. This protocol showed promising results in internal benchmarking, and we expect to achieve higher accuracy in this important class of polypeptide simulations using the new force-field. The long term (Phase II) goal of this proposal is to further expand ECEPP05 into a general-purpose force-field that could be applied to small molecules and in particular in docking. Other extensions will include parameters for DNA and metalloprotein modeling. PUBLIC HEALTH RELEVANCE: Molecular mechanics techniques increasingly allow in-silico simulation of essential biological processes at the atomic level. Molsoft's ICM (Internal Coordinate Mechanics) software platform is a particularly efficient modeling tool because of its use of internal variables, rather than traditional Cartesian coordinates, in the description of the molecular structure. ICM is increasingly used by researchers in various structural and computational biology projects, including computer-assisted drug discovery. The force-fields that quantitatively describe atomic interactions are at the heart of molecular mechanics modeling, and the accuracy of molecular simulation can only be as good as the underlying force-fields. Equipping structural biologists with the latest ECEPP05 force-field from Cornell University, together with its further improvements at Molsoft, will help bring molecular simulations and structure predictions to the next level of accuracy. Successful structure modeling will ultimately help improve our understanding of molecular mechanisms underlying disease and help accelerate structure-based drug discovery efforts.