CRCNS US-German Research Proposal: Combining computational modeling and artificial intelligence to understand receptor function in physiology and disease

g蛋白偶联受体（gpcr）是细胞表面受体，将外部信号（如药物结合）转化为细胞内过程的激活。受体的三维形状取决于与其结合的分子和外部环境（例如受伤组织与正常组织）。为了对GPCR的所有可能构象进行采样，使用了一种强大的计算模拟方法，将传统的分子动力学模拟与人工智能相结合。这种新方法被应用于阿片受体，一个突出的gpcr亚家族。这些受体在受伤和发炎组织中介导疼痛缓解，但在健康组织中也有不良副作用，如呼吸抑制或大脑镇静。由于发炎的环境，阿片受体改变了它们的构象。如果有选择性地针对这种“病理”形式的阿片受体的药物，它们将在不影响健康环境中的受体的情况下治疗急性疼痛，从而避免传统阿片类药物所观察到的不良副作用。这种方法可以在未来用于发现只影响受伤组织中的阿片受体的更安全的止痛药。分子动力学模拟与人工智能的创新结合使大量阿片受体构象的硅采样成为可能。有效的模拟将利用最近开发的人工神经网络方法，有效地表示高维分子相互作用的动力学。相应的数学理论并不局限于分子模拟；原则上，它可以应用于任何生成的马尔可夫过程。分子模拟将提供对环境因素（如pH值和自由基的存在）的影响的深入了解，并将用于实验室实验的建议。计算模型将在实验室中使用阿片受体突变体和抗体进行测试，这些抗体可以将受体锁定在活跃状态。利用蛋白质设计和酵母展示技术的结合，将产生模仿g蛋白亚基的抗体和微型蛋白。一旦这些方法在阿片受体上建立起来，它们可能会扩展到参与其他神经系统疾病的gpcr。最终，新的计算方法与体外实验的结合将使GPCR信号在健康与受伤环境中的系统研究成为可能。德国联邦教育和研究部（BMBF）正在资助一个伙伴项目。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。

G-protein coupled receptors (GPCRs) are cell surface receptors that translate external signals (e.g. binding of drugs) into the activation of processes inside the cell. The three-dimensional shape of the receptor changes depending on the molecule that is binding to it and the external environment (e.g. injured versus normal tissue). To sample all possible conformations of a GPCR, a powerful computational simulation approach is used that combines traditional molecular dynamics simulations with artificial intelligence. This novel approach is applied to opioid receptors, a prominent subfamily of GPCRs. These receptors mediate pain relief in injured and inflamed tissue, but also have adverse side effects in healthy tissue, such as depression of breathing or sedation in the brain. The opioid receptors change their conformation as a result of the inflamed environment. If there were drugs that selectively targeted this "pathological" form of opioid receptors, they would treat acute pain without affecting receptors in a healthy environment, and thereby avoid the adverse side effects observed for conventional opioids. This approach can be used in the future to discover safer pain killers that only affect opioid receptors in injured tissues. The innovative combination of molecular dynamics simulations with artificial intelligence enables the sampling in silico of large numbers of opioid receptor conformations. Efficient simulation will take advantage of a recently developed artificial neural network approach that efficiently represents the dynamics of high-dimensional molecular interactions. The corresponding mathematical theory is not restricted to molecular simulation; in principle, it could apply to any generated Markov process. The molecular simulations will provide insight into the effects of environmental factors such as pH and the presence of free radicals and will be used to suggest experiments in the laboratory. Computational models will be tested in the lab using opioid receptor mutants and antibodies that can lock the receptors into an active state. Antibodies and miniproteins mimicking G-protein subunits will be generated using a combination of protein design and yeast display technologies. Once these methods have been established for opioid receptors, they may be extended to GPCRs involved in other nervous system disorders. Ultimately, the combination of novel computational methods with in vitro experiments will enable a systematic study of GPCR signaling in healthy versus injured environments.A companion project is being funded by the Federal Ministry of Education and Research, Germany (BMBF).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.