Hybrid Computational Models for Membrane-Protein Interfaces

在化学系化学理论、模型和计算方法项目的支持下，波士顿大学崔强教授正在开发有效的计算方法来解决涉及脂质膜界面上蛋白质/肽集体行为的问题。由于涉及多个长度和时间尺度，这些问题很难用现有的计算方法来研究。利用机器学习（ML）技术的最新进展，崔博士将致力于克服这些挑战，建立高效可靠的计算模型，从而能够对细胞膜表面和蛋白质介导的膜孔处的蛋白质相分离进行机制分析，这在细胞信号传导、病毒感染和突触传递等重要生物过程中至关重要。崔还将参与各种教育和外展活动，以激励具有广泛背景的学生在物理化学，计算科学和生物学之间的边界上追求职业。在本科阶段，崔教授将努力加强波士顿大学化学课程中计算和基本编程概念的整合。为了实现研究目标，崔团队将在独特的生物物理背景下有效地整合模拟方法的最新进展。在其中一个问题中，Cui及其同事将致力于了解在液-液相分离的情况下蛋白质-膜相互作用如何改变蛋白质的构象和相互作用特性。独特的角度将是开发一种混合ML/MM模型，其中蛋白质及其与膜环境的相互作用使用ML来描述，用原子模拟数据和参考粗粒度模型进行训练；这种混合模型的优点是在粗粒度水平上捕获了多体效应，这一特征对于正确描述不同环境下蛋白质的集体行为至关重要，包括脂质膜的相分离或润湿。在另一个问题中，挑战在于理解多种多肽或蛋白质基序调节膜孔的机制。Cui小组将结合有限温度串和ML方法来自动系统地扩展集体变量列表，以评估潜在的最小自由能途径。结合全局（字符串）和局部（PIB）增强采样技术的优势的基本策略将潜在地适用于广泛的问题，在这些问题中，一组最小的全局进度变量是提前已知的，但重要的局部自由度仍然模糊不清。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。

WIth support from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry, Professor Qiang Cui of Boston University is developing effective computational methods for tackling problems that involve collective behaviors of proteins/peptides at the lipid membrane interface. These problems are difficult to study using existing computational methodologies due to the involvement of multiple length and time scales. Taking advantage of recent progress in machine learning (ML) techniques, Dr. Cui will aim to overcome these challenges to establish efficient and reliable computational models that enable the mechanistic analysis of protein phase separation at cell membrane surface and protein mediated membrane porations, which are critical in important biological processes such as cell signaling, viral infection and synaptic transmission. Cui will also engage in various education and out-reach activities to inspire students of broad backgrounds to pursue a career at the boundary between physical chemistry, computational science, and biology. At the undergraduate level, Professor Cui will endeavor to enhance the integration of computation and basic programming concepts into the chemistry curriculum at Boston University.To accomplish the research goals, the Cui team will effectively integrate recent advances in simulation methodologies in unique biophysical contexts. In one problem, Cui and co-workers will aim to understand how protein-membrane interactions modify the conformational and interaction properties of proteins in the context of liquid-liquid phase separation. The unique angle will be to develop a hybrid ML/MM model in which the protein and its interaction with the membrane environment are described using ML, trained with atomistic simulation data and a reference coarse-grained model; the advantage of this hybrid model is that many-body effects at the coarse-grained level are captured, a feature expected to be essential to the proper description of collective behaviors of proteins in different environments, including phase separation at or wetting of the lipid membrane. In another problem, the challenge is to understand the mechanism by which multiple types of peptides or protein motifs regulate membrane pores. The Cui group will combine finite temperature string and an ML approach to expand the list of collective variables automatically and systematically for evaluating the underlying minimum free energy pathways. The fundamental strategy of combining the strengths of global (string) and local (PIB) enhanced sampling techniques will be potentially applicable to a broad range of problems in which a minimal set of global progress variables is known ahead of time, yet important local degrees of freedom remain obscure.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.