Absolute binding free energies for virtual screening: A novel implementation of quantum mechanics/molecular mechanics (QM/MM) for FEP that allows substantial sampling and a significant quantum region

用于虚拟筛选的绝对结合自由能：用于 FEP 的量子力学/分子力学 (QM/MM) 的新颖实现，允许大量采样和重要的量子区域

• 批准号: 10759829
• 负责人: David A Pearlman
• 金额: $ 27.34万
• 依托单位: QUANTUM SIMULATION TECHNOLOGIES, INC.
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-01 至 2024-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10759829
• 关键词: Address; Adoption; Binding; CDK2 gene; Computers; Computing Methodologies; Data; Databases; Docking; Drug Targeting; Evaluation; Explosion; Free Energy; Goals; Head; Lead; Ligand Binding; Ligands; Mechanics; Methods; Modification; Molecular; Output; Process; Proteins; Protocols documentation; Publications; Quantum Mechanics; Reliability of Results; Resources; Sampling; Savings; Speed; System; Tail; Testing; Thrombin; Time; Triage; Virtual Tool; Work; beta-site APP cleaving enzyme 1; computational chemistry; cost; data modeling; drug candidate; drug discovery; falls; improved; interest; lead candidate; lead optimization; mechanical force; molecular mechanics; novel; novel strategies; programs; quantum; receptor; restraint; screening; simulation; timeline; virtual; virtual screening

Project Summary Computational chemistry has revolutionized drug discovery, reducing by months or even years the amount of time it takes to discover and refine a lead candidate. Nowhere has the contribution of computational chemistry been greater than in the realm of virtual screening (VS) to identify an initial hit to a drug target receptor. It is now routine to screen 106-108 virtual compounds via molecular docking to identify potential binders. A small number of these will be purchased and screened, which is a slower and more expensive process. While docking is demonstrably useful for brute force triage, it is also generally unreliable for rank-ordering the compounds that survive the triage. There is a substantial and unmet need for computational methods that are better at rank ordering that can further reduce the number of compounds that survive to purchase/screening. Interest is growing in an approach termed ABFE, in which the Absolute Binding Free Energies of diverse ligands can be evaluated for a common protein receptor target. This approach is a natural outgrowth of relative binding free energy (RBFE) methods, the most well-known of which is Free Energy Perturbation (FEP). In recent years, the application of FEP for hit-to-lead optimization has exploded, thanks to increasing computer resources and automized workflows. With a reliable ABFE approach, further enrichment of the potential binders that come from dock-based screening can be obtained, improving the cost/hit ratio for the expensive experimental tail of the screening campaign. Based on FEP, the computational formalism that would make ABFE calculations possible within the screening paradigm has been described, and a few publications have demonstrated that it is, indeed, capable of further enriching the compounds that survive the molecular docking screen. These calculations have still been limited by two issues: 1) Computational throughput; 2) Limitations of the Molecular Mechanics (MM) force field that has been exclusively used in these ABFE/FEP simulations. The limitations of computational throughput are increasingly addressed by expansion in the availability of cloud resources, so the limitations of the MM force field are the primary issue. We propose an ABFE/FEP approach that replaces the limited MM representation with one based on a combined quantum mechanics (QM) +MM approach: QM/MM—where the region of ligand binding is treated using QM. In contrast to MM, QM describes molecular energetics much more exactly, and is broadly appliable to all classes of molecular ligands, unlike MM, which has a large number of known limitations/deficiencies. We will apply ABFE/FEP calculations to a variety of systems to validate the approach within the context of virtual screening, and to demonstrate the improvements that QM/MM allows versus traditional MM approaches.

项目概要 计算化学彻底改变了药物发现，将药物的用量减少了数月甚至数年。 发现和完善主要候选者需要时间，没有计算化学的贡献。 比虚拟筛选 (VS) 领域更能识别药物靶标受体的初始命中。 现在常规通过分子对接筛选 106-108 个虚拟化合物，以识别潜在的 A 小 A 结合物。 其中一些将被购买和筛选，这是一个更慢且更昂贵的过程。 对接对于强力分类显然很有用，但对于排序排序通常也是不可靠的 在分类中幸存下来的化合物对计算方法存在大量且未得到满足的需求。 更好地进行排序，可以进一步减少购买/筛选中幸存的化合物数量。 人们对一种称为 ABFE 的方法越来越感兴趣，其中不同物质的绝对结合自由能 可以评估配体的常见蛋白质受体靶标，这种方法是相对的自然产物。 结合自由能（RBFE）方法，其中最著名的是自由能扰动（FEP）。 近年来，由于计算机的不断增加，FEP 在从命中到先导优化方面的应用呈爆炸式增长 资源和自动化工作流程。 通过可靠的 ABFE 方法，进一步丰富来自码头的潜在粘合剂 可以获得筛选，提高筛选昂贵的实验尾部的成本/命中率 活动。 基于 FEP 的计算形式使 ABFE 计算在筛选中成为可能 范式已经被描述，并且一些出版物已经证明它确实能够进一步 富集分子对接筛选中幸存的化合物这些计算仍然有限。 两个问题：1）计算吞吐量；2）分子力学（MM）力场的局限性 已专门用于这些 ABFE/FEP 模拟，但计算吞吐量的限制是。 云资源可用性的扩展越来越多地解决了 MM 力量的局限性 领域是首要问题。 我们提出了一种 ABFE/FEP 方法，用基于 组合量子力学 (QM) +MM 方法：QM/MM——处理配体结合区域 使用 QM 与 MM 相比，QM 更准确地描述分子能量学，并且适用范围广泛。 适用于所有类别的分子配体，与 MM 不同，MM 具有大量已知的限制/缺陷。 我们将把 ABFE/FEP 计算应用于各种系统，以在以下环境中验证该方法： 虚拟筛选，并展示 QM/MM 相对于传统 MM 方法的改进。