Tackling Multifaceted Drug Design Problems with Lambda Dynamics Based Technologies

项目摘要 现代药物发现是一个漫长而昂贵的过程，需要科学家团队，多年的研究， 研究和数百万美元，以确定临床前候选药物适合临床试验。掺入 将计算工具应用于药物发现已被证明是降低这些成本的有效手段。全原子 分子动力学模拟与炼金术自由能计算相结合， 用于研究蛋白质-配体复合物的结构和热力学性质以及优化药物的工具 用于改善与感兴趣的靶的结合亲和力的候选物。Lambda dynamics（LD），一个新的炼金术免费 能量方法，便于采样的多个扰动的化学系统，同时，在 单分子动力学模拟，克服了与传统方法相关的固有可扩展性限制， 自由能方法迄今为止，各种化学扰动，包括不同的配体官能团， 转化和蛋白质侧链突变，已经用（LD）在单个化学实体上进行， 例如，在一个实施例中，一种小分子或蛋白质，并取得了很大的成功。几十到几百种化学状态已经被有效地 与传统方法相比，使用数量级更少的计算资源进行采样。这 一项提案寻求支持，以这些发现为基础，并应用基于LD的技术来探索多方面的 药物发现中的设计问题，其特征在于对多个结合伴侣进行化学修饰。具体地说， 将研究药物发现的三个具有挑战性的领域：（1）了解和克服耐药性 源自药物靶标中的错义突变，（2）表征蛋白质-蛋白质相互作用和结合 特异性，和（3）通过整合新的靶特异性先导化合物类似物的自动生成 使用机器或深度学习算法进行LD计算。这些努力的成功将需要寻找 通过大的组合化学空间，这只能用基于LD的技术来实现。模型 对于多发性骨髓瘤或阿尔茨海默病具有高度治疗重要性的蛋白质靶向系统将被 为实现我们的目标而努力。因此，这项工作将有助于加速基于临床前结构的 药物设计，使复杂的分子设计方案，以解决这些毁灭性的疾病。

Project Summary Modern day drug discovery is a long and expensive process requiring teams of scientists, multiple years of research, and millions of dollars to identify preclinical drug candidates suitable for clinical tests. The incorporation of computational tools into drug discovery has proved an effective means to reduce these costs. All-atom molecular dynamics simulations coupled with alchemical free energy calculations have been extremely beneficial tools for studying structural and thermodynamic properties of protein-ligand complexes and optimizing drug candidates for improved binding affinity to a target of interest. Lambda dynamics (LD), a newer alchemical free energy method, facilitates the sampling of multiple perturbations to a chemical system, simultaneously, within a single molecular dynamics simulation, overcoming inherent scalability limitations associated with conventional free energy methods. To date, a variety of chemical perturbations, including diverse ligand functional group transformations and protein side chain mutations, have been performed with (LD) on a single chemical entity, e.g., a small molecule or protein, with much success. Tens to hundreds of chemical states have been efficiently sampled using an order of magnitude less computational resources compared to conventional methods. This proposal seeks support to build upon these findings and apply LD-based techniques to explore multifaceted design problems in drug discovery featuring chemical modifications on multiple binding partners. Specifically, three challenging areas of drug discovery will be investigated: (1) understanding and overcoming drug resistance originating from missense mutations in a drug target, (2) characterizing protein-protein interactions and binding specificities, and (3) automating the generation of novel, target-specific lead compound analogs by integrating LD calculations with machine- or deep-learning algorithms. Success in these efforts will require searching through large combinatorial chemical spaces that can only be accomplished with LD-based techniques. Model protein-target systems of high therapeutic importance from Multiple Myeloma or Alzheimer’s Disease will be investigated in accomplishing our goals. Thus, this work will assist in accelerating preclinical structure-based drug design by enabling complex molecular design scenarios to be addressed in these devastating diseases.