Kinetics-Driven Drug Discovery Using Persistent Homology, Rare-Event Molecular Dynamics and Experimental Data

在这个项目中，一组来自数学、分子生物学和药物化学的研究人员将开发数学和计算工具来预测可能有助于治疗糖尿病患者神经性疼痛的化合物的功效。药物大多由非常小的分子组成，这些分子通过与我们体内的目标生物分子结合并干扰其功能来发挥作用。设计新药——比如治疗癌症、糖尿病和阿尔茨海默病的药物——面临的一个主要挑战是弄清楚如何既准确地结合特定的靶标（几乎没有脱靶结合），又有效地结合（药物分子占据靶标的高比例）。最大限度地提高药物有效性的一个关键指标是它的“停留时间”，即每次结合事件后药物在结合位点停留的平均时间。然而，我们对药物分子的结构如何决定其停留时间知之甚少，这阻碍了我们在药物设计过程中结合停留时间预测的能力。这项研究将预测化合物与影响糖尿病神经性疼痛的药物靶分子结合的停留时间，并通过实验对这些预测进行测试。这项研究可能会导致糖尿病神经性疼痛的新治疗方法，也可以作为未来药物发现工作的蓝图，重点是停留时间。为了促进采用这种方法，研究小组将通过专门的网站、在线服务器和参加预测药物结合特性的全球竞赛来传播他们的结果。该项目还包括培养具有独特跨学科背景的研究生，并将为数学和生物科学交叉的研究生课程的发展提供信息。该项目将开发一系列数学和计算工具，以实现基于动力学的药物发现。这些研究将针对可溶性环氧化物水解酶（sEH）进行，sEH是糖尿病神经性疼痛的既定药物靶点，目前只有有限的药物获得批准。该项目将采用综合方法，包括拓扑建模、机器学习、虚拟筛选、分子模拟以及体外和体内化合物功效评估。在皮伟的实验室中，持续同源性将与深度学习结合使用，从蛋白质配体复合物中提取拓扑信息，并预测稳定的结合姿态、结合亲和力和结合动力学。人们相信，拓扑分析和深度学习的结合将具有变革性：在不久的将来，它将在3D生物分子数据预测中带来类似方法的激增，以及在化学和材料科学等其他领域的应用。PI Dickson将使用分子建模中的稀有事件技术来模拟配体释放事件，并确定配体结合和释放过程的限速过渡状态。该项目还将首次研究配体结合过渡态的鲁棒性，这是实现基于动力学的药物设计的关键数量。第三，PI Lee将持续评估预测化合物的结合亲和力和停留时间。选定的化合物将在体内用一种新的小鼠模型进行测试，以确定长时间体外停留时间的益处。这个合作项目将通过汇集高等数学、计算生物物理学和分子药理学的专业知识来实现协同效益。从这项工作中产生的采样和预测的协作工具可以应用于发现其他感兴趣目标的新型长停留时间化合物。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In this project a team of investigators from mathematics, molecular biology and medicinal chemistry will develop mathematical and computational tools to predict the efficacy of compounds that may help treat neuropathic pain in diabetics. Pharmaceutical drugs are mostly made up of very small molecules, which take their effect by binding to target biomolecules in our bodies and perturbing their functions. A major challenge in designing new drugs - such as treatments for cancer, diabetes and Alzheimer's disease - is figuring out how to bind a particular target both accurately (with little off-target binding), and effectively (a high percentage of targets occupied by drug molecules). A key quantity to maximize the effectiveness of a drug is its "residence time," the average amount of time the drug will remain in the binding site after each binding event. However, little is known about how the structure of a drug molecule determines its residence time, and this hinders our ability to incorporate residence time predictions in the drug design process. This research will predict the residence times of compounds binding to a pharmaceutical target molecule that affects diabetic neuropathic pain, as well as test those predictions experimentally. This study will potentially result in new treatments for diabetic neuropathic pain and also serve as a blueprint for future drug discovery efforts focused on residence time. To facilitate adoption of this approach the team of investigators will disseminate their results via a dedicated website, online servers and participation in world-wide competitions for predicting drug binding properties. This project also involves the training of graduate students with unique interdisciplinary backgrounds, and will inform the development of graduate courses at the intersection of mathematics and biological sciences. This project will develop a pipeline of mathematical and computational tools to enable kinetics-based drug discovery. The studies will be conducted on soluble epoxide hydrolase (sEH), an established pharmaceutical target for diabetic neuropathic pain for which only limited drugs have yet been approved. This project will use an integrated approach that encompasses topological modeling, machine learning, virtual screening, molecular simulation, as well as in vitro and in vivo assessment of compound efficacy. In PI Wei's laboratory, persistent homology will be used together with deep learning to abstract topological information from protein-ligand complexes and predict stable binding poses, binding affinities, and binding kinetics. It is believed that the combination of topological analysis and deep learning will be transformative: it will bring a surge in similar approaches in 3D biomolecular data predictions in the near future, as well as applications to other fields, such as chemistry, and material science. PI Dickson will use rare-event techniques in molecular modeling to simulate ligand release events, and identify the rate-limiting transition states of the ligand binding and release process. This project will also examine the robustness of ligand binding transition states for the first time, which is the key quantity to enable kinetics-based drug design. Thirdly, PI Lee will continually assess the binding affinity and residence times of the predicted compounds. Selected compounds will be tested in vivo with a novel mouse model, to determine the limits of the benefits of long in vitro residence times. This collaborative project will achieve synergistic benefits by bringing together expertise from advanced mathematics, computational biophysics, and molecular pharmacology. The collaborative tools for sampling and prediction resulting from this work can then be applied to the discovery of novel long residence time compounds for other targets of interest.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.