RUI: D3SC: Advancement of Excitation Energy Transfer Modeling with Deep Learning Algorithms

蒙茅斯大学的Dmitri Kosenkov获得了化学系化学理论、模型和计算方法项目的奖励，通过结合深度学习技术，推进了复杂分子系统中电子激发能量传递动力学建模的计算方法。激发能传递是光在自然和人工系统中收集和利用的重要过程。现代技术，如光电和分子传感器，涉及能量转移。特别是，了解分子荧光探针中的激发能转移机制对于靶向蛋白质、DNA序列和感知细胞内环境至关重要。分子系统的大尺寸和复杂结构给激发能传递过程建模带来了挑战。科森科夫教授及其同事正在将机器学习与计算化学技术相结合，以克服这些障碍。这种方法可以加快计算速度，降低计算成本。这些科学知识在化学生物学、人工光合作用、有机光伏和光疗法等领域有很大的发展潜力。科森科夫教授的研究涉及以本科生为主的蒙茅斯大学（Monmouth University）的本科生。Kosenkov小组通过计算化学软件和基于研究的实验室项目向更广泛的科学界传播计算方法，这些项目可以在美国和世界各地的学院和大学中实施。最后，这项工作支持蒙茅斯大学为本科生和高中生举办的计算化学研讨会，在那里，计算化学、化学生物学和机器学习领域的顶尖科学家分享他们的知识和经验。这项研究的重点是新型的生物正交硼二吡咯烯衍生探针，它为细胞器的可视化、细胞内药物分布提供了一种革命性的工具，并为在光激活下通过控制产生细胞毒性单线态氧的潜在光疗提供了可能性。激发能转移是这些分子荧光探针工作的关键过程。科森科夫教授和他的合作者正在开发一种新颖的计算方法，将量子动力学理论与深度学习算法相结合。利用量子主方程方法对能量传递动力学进行了建模。深度学习组件能够高通量筛选电子激发能和耦合，进一步用于量子动力学模拟。该方法的优势在于其灵活性和适用于广泛的分子系统。这些方法能够阐明和比较不同化学结构的分子荧光探针的激发能转移机制。此外，软件工具模拟了广泛的分子系统，并可能允许随着领域的发展进一步迭代改进。教育部分的重点是开发和传播一系列以研究为基础的多周计算实验单元，主要用于物理和计算化学课程。这些实验室单元涵盖了深度学习算法的基础知识及其在化学中的应用，包括荧光探针的建模。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。

Dmitri Kosenkov of Monmouth University is supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to advance computational methods for modeling electronic excitation energy transfer dynamics in complex molecular systems by the incorporation of deep learning techniques. Excitation energy transfer is a fundamentally important process for light harvesting and utilization in natural and artificial systems. Modern technologies, such as photovoltaics and molecular sensors, involve energy transfer. In particular, understanding excitation energy transfer mechanisms in molecular fluorescent probes is crucial for targeting proteins, DNA sequences, and sensing of intracellular environments. The challenges in modeling excitation energy transfer processes is associated with the large sizes and complex structures of the molecular systems. Professor Kosenkov and coworkers are incorporating machine learning into computational chemistry techniques to overcome these obstacles. This approach may accelerate computations and reduce computational cost. The scientific knowledge has the potential to advance in the fields of chemical biology, artificial photosynthesis, organic photovoltaics, and phototherapies. Professor Kosenkov involves undergraduate students at Monmouth University, a primarily undergraduate institution, in his research. The Kosenkov group disseminates the computational methods to the broader scientific community via computational chemistry software and research-based laboratory projects that can be implemented in colleges and universities in the United States and worldwide. Finally, the work supports computational chemistry workshops for undergraduate and high school students at Monmouth University, where leading scientists in the fields of computational chemistry, chemical biology, and machine learning share their knowledge and experience. The research is focused on novel bioorthogonal boron dipyrromethene-derived probes that offer a revolutionary tool for visualization of cell organelles, intracellular drug distribution, and open a possibility for potential phototherapy through controlled generation of cytotoxic singlet oxygen under light activation. Excitation energy transfer is a key process for operation of these molecular fluorescent probes. Professor Kosenkov and his collaborators are developing an original computational methodology that integrates quantum dynamics theories with deep learning algorithms. The dynamics of energy transfer is modeled with the quantum master equation methods. The deep learning component enables high-throughput screening of electronic excitation energies and couplings further used in quantum dynamics simulations. The methodology is advantageous due to its flexibility and applicability to a wide range of molecular systems. The methods enable elucidation and comparison of the mechanisms of excitation energy transfer for molecular fluorescent probes of varied chemical structures. Moreover, the software tools simulate a wide range of molecular systems and may allow for further iterative improvements as the field develops. The educational component is focused on development and dissemination of a sequence of research-based multi-week computational laboratory units, primarily for physical and computational chemistry courses. These laboratory units cover basics of deep leaning algorithms and their applications in chemistry including modeling of fluorescent probes.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.