Next-generation integrated quantum force fields for biomedical applications

用于生物医学应用的下一代集成量子力场 PI：Darrin M.约克，罗格斯大学，皮斯卡特维，新泽西州08854-8087美国。 我们最近开发了下一代量子力学力场（QMFFs）的新框架 旨在满足生物分子模拟和药物发现应用的挑战。QMFF具有 巨大的计算优势，相对于他们的完全QM同行，是固有的并行化， 线性缩放，提供了巨大的计算加速，并有希望的定量准确性潜在的 上级于全面质量管理方法。QMFF精确地模拟多极静电，电荷渗透效应， 非线性偏振响应因此，QMFF为药物发现应用提供了一种变革性技术， 特别是，用于提高铅精炼中自由能模拟的预测能力。这些都是至关重要的 重要的是药物分子的不同化学空间，包括卤素键合，阳离子和金属配体 交互.此外，QMFF提供了用于模拟共价抑制剂的机制。具体而言，我们建议： 开发新的QMFF用于药物发现。QMFF将基于半经验和从头算两种方法开发 密度泛函方法分以下几个阶段：1）确定多极映射参数， DFTB静电势达到更高的精度，2）电子响应项的增加， 用正交扰动-响应法求解化学势平衡（CPE）修正 非静电非键相互作用的参数化， 使用捕获多体交换和分散相互作用的现实势能的作用参数，以及4） 探索统计潜力，使用应用于量子数据集的机器学习方法， 内部构象能和短程相互作用。二.开发新的自由能源方法， 使用QMFF的蛋白质-配体结合预测。我们将开发一种新型的综合免费能源管道， 配体和抑制剂的dict炼金术结合自由能。这将包括新的GPU加速方法 空间自适应混合采样（-SAMS）和2D-vFEP分析，结合构象空间 增强的采样方法的炼金术步骤的热力学循环，和进步的自由en- 能量“图书终结”方法（BBQm），有效地连接分子力学力场和QMFF模型 表示。三.测试和验证QMFF和自由能方法，并适用于MIF抑制剂结合。的 方法将广泛测试对已建立的数据集的溶剂化自由能，和药物的发现 数据集将通过检查以下物质的相对结合自由能进行更深入的验证研究： 巨噬细胞抑制因子（MIF）的抑制剂。最后，探索性应用程序将检查机制， 表征过渡态并预测一系列MIF抑制剂的共价抑制率。

Next-generation integrated quantum force fields for biomedical applications PI: Darrin M. York, Rutgers University, Piscataway, NJ 08854-8087 USA. We have recently developed novel framework for next-generation quantum mechanical force fields (QMFFs) designed to meet the challenges of biomolecular simulations and drug discovery applications. QMFFs have tremendous computational advantages relative to their fully QM counterparts, being inherently parallelizable and linearly scaling, offering tremendous computational speedup, and promising quantitative accuracy potentially superior to full QM methods. QMFFs accurately model multipolar electrostatics, charge penetration effects, and non-linear polarization response. QMFFs thus offer a transformative technology for drug discovery applications, in particular, for advancing the predictive capability of free energy simulations in lead refinement. These are critically important for the diverse chemical space of drug molecules, including halogen bonding, cation- and metal-ligand interactions. Further, QMFFs offer a mechanism for modeling covalent inhibitors. Specifically, we propose to: I. Develop new QMFFs for drug discovery. QMFFs will be developed based on both semiempirical and ab initio density-functional methods in the following stages: 1) determination of multipolar mapping parameters enhancing the DFTB electrostatic potential to reach greater accuracy, 2) augmentation of electronic response terms using chemical potential equalization (CPE) corrections using an orthogonal perturbation-response approach to solve the under-polarization problem of DFTB methods, 3) parameterization of non-electrostatic non-bonded interac- tion parameters using realistic potentials that capture many-body exchange and dispersion interactions, and 4) exploration of statistical potentials, using machine learning approaches applied to quantum data sets, to correct internal conformational energies and short-range interactions. II. Develop new free energy methods to enable protein-ligand binding predictions using QMFFs. We will develop a novel integrated free energy pipeline to pre- dict alchemical binding free energies for ligands and inhibitors. This will include new GPU-accelerated methods for -space self-adaptive mixture sampling ( -SAMS) and 2D-vFEP analysis, coupled with conformational space enhanced sampling methods for alchemical steps of the thermodynamic cycle, and advancements in free en- ergy “book-ending” methods (BBQm) to efficiently connect molecular mechanical force field and QMFF model representations. III. Test and validate QMFFs and free energy methods, and apply to MIF inhibitor binding. The methods will be broadly tested against established data sets for solvation free energies, and a drug discovery data set. More in-depth validation studies will be conducted by examining the relative binding free energies of inhibitors of the macrophage inhibitory factor (MIF). Finally, exploratory applications will examine mechanisms, characterize transition states and predict rates for covalent inhibition for a series of MIF inhibitors.