Develop light-oxygen-voltage (LOV) sensing optogenetics tools through novel computational approaches with experimental validation

摘要 光遗传学是一种强大的技术，它综合了光(光学)和基因工程的使用。光- 氧-电压(LOV)结构域是光响应的昼夜节律时钟调节蛋白，是一种新的 为光遗传工具开发提供了平台。描述了不同LOV结构域的变构机制 对这样的发展至关重要。分子动力学(MD)模拟是揭示 变构机制作为原子水平上的时空信息。然而，有两条主干道-- 阻止当前可用的MD模拟方法来阐明LOV结构域机制：1)时间有限 规模；2)缺乏动态信息。许多改进的抽样方法被开发出来，以隐含地增加 动力学模拟的可用时间尺度，但不适用于蛋白质变构机理的模拟。 由于需要先验地构造反应坐标而导致的缺陷。为了解决这个问题，我们最近 应用深度学习方法--自动编码法，建立了一种新的基于自适应编码的数据降维模型 类固醇蛋白。这些模型的主要优势是能够准确地再生蛋白质三级 来自低维空间的结构，也称为。潜伏空间。目前还缺乏蛋白质的动力学模型。 变构作用下的构象变化。我们建立了一个定向动力学转移网络(DKTN)模型 在之前的获奖期内，根据分子动力学模拟对蛋白质构象变化的动力学进行了建模。 基于我们最近的工作，在这个应用中，我们将继续开发两种新的方法，自动编码 基于潜在空间(AELS)采样方法和机器学习的有向动力学转移网络(ML- DKTN)方法，并应用这些新方法来阐明LOV结构域的机制。有了这些经验- 为了进行心理验证，我们将进一步开发选定LOV结构域的关键突变体，作为新的光遗传工具。 我们期望开发一套有效的计算工具来获得与变构函数相关的构象 系综和动力学模型。我们将应用这些工具来构建构象系综和动力学模型 关键的LOV结构域蛋白及其突变体描述其潜在的变构机制。这些是- 理论模型可以直接指导光遗传工具的进一步发展。 选定的LOV域。在拟议的研究中确定的有希望的突变体将接受实验 通过生物物理表征进行验证。拟议的研究活动还将提供独特的培训- 为具有不同背景的本科生和研究生提供帮助的ING活动 科学研究，提高他们的研究、人际交往和沟通能力。

SUMMARY Optogenetics is a powerful technique that integrates the use of light (optics) and genetic engineering. Light- oxygen-voltage (LOV) domains are light-responsive circadian clock regulation proteins and serve as a novel platform for the optogenetic tools development. Delineating allosteric mechanisms of various LOV domains is critical for such development. Molecular dynamics (MD) simulations are the main computational tools to reveal allosteric mechanisms as spatial-temporal information at the atomic level. However, there are two major road- blocks to the currently available MD simulation methods to elucidate LOV domain mechanisms: 1) limited time scale; 2) lack of kinetic information. Many enhanced sampling methods were developed to implicitly increase the accessible time scale of dynamics simulations, but are not suitable for simulations of protein allosteric mecha- nisms due to the requirement of constructing reaction coordinates a priori. To address this issue, we recently applied deep learning methods, named autoencoders, to develop novel dimensionality reduction models for al- losteric proteins. The main advantage of these models is the ability to accurately regenerate protein tertiary structure from the low dimensional space, a.k.a. latent space. There is also a lack of kinetics models for protein conformational changes underlying allostery. We developed a directed kinetic transition network (DKTN) model during the previous award period to model kinetics of protein conformational change based on MD simulations. Based on our recent work, in this application, we will continue to develop two new methods, the auto-encoded latent space (AELS) sampling methods and machine learning based directed kinetic transition network (ML- DKTN) methods, and apply these novel methods to elucidate the LOV domain mechanisms. With the experi- mental validation, we will further develop key mutants of the selected LOV domains as new optogenetic tools. We expect to develop a set of efficient computational tools to obtain allosteric function-related conformational ensembles and kinetics models. We will apply these tools to build conformational ensembles and kinetics models for key LOV domains proteins and their mutants to delineate their underlying allosteric mechanisms. These the- oretical models could provide direct guidance for the further development of optogenetic tools based on the selected LOV domains. The promising mutants identified in the proposed study will be subjected to experimental verification through biophysical characterization. The proposed research activities will also provide unique train- ing activities for motivated undergraduate and graduate students with various backgrounds to contribute to scientific research and improve their research, interpersonal, and communication skills.