Development of Coarse-Grained Models and Computational Approaches for Studying Structure in Solutions of Cellulose Derivatives

纤维素是一种丰富的天然生物聚合物，甲基纤维素是通过对纤维素进行无毒化学取代而获得的。导致纤维素形成甲基纤维素的化学过程破坏了导致纤维素在水中不溶解的分子水平的相互作用。甲基纤维素在水中溶解度的提高及其丰富的天然原料（纤维素）使得甲基纤维素水溶液在许多应用中都很有用，如食品添加剂、油漆去除剂、粘合剂、乳化剂和可生物降解的包装材料。为了使甲基纤维素水溶液的物理性质适合上述应用，需要进行基础研究，以了解甲基纤维素水溶液和凝胶的物理性质随温度和浓度的变化，以及驱动这些性质的甲基纤维素链的潜在分子结构。与真实实验相比，分子模拟既便宜又有效，而且可以作为有价值的微观工具，提供对聚合物溶液结构的分子洞察。然而，对于甲基纤维素溶液和凝胶，只有少数计算研究，部分原因是这些材料的复杂性，部分原因是缺乏良好的分子模型。该项目旨在开发更好的甲基纤维素模型和计算方法，以便在不同温度和浓度的水中对甲基纤维素链的结构进行基础研究，并指导甲基纤维素溶液在各种日常应用中的实际使用。PI将把模型和计算方法的开发整合到她在特拉华大学的跨学科分子建模和软材料模拟选修课程中。本课程每两年开设一次，面向（化学与材料）工程和物理科学（物理、化学）专业的本科生和研究生。该项目还将包括该项目在化学工程本科课程“化学工程师概率与统计”中的数据科学方面。为了更好地招募和留住计算材料领域的女科学家，PI Jayaraman将继续组织研讨会/讲座，就像她在2020-21年发起的成功的“WELCOME: women excellence in computational Molecular Engineering”虚拟月度系列研讨会一样。这种虚拟研讨会将继续为研究界的女研究生和早期职业研究人员提供交流机会，并帮助招聘和留住从事STEM职业的妇女和研究人员。技术概述：该PI建议开发新的粗粒度（CG）模型和计算方法，以了解甲基纤维素水溶液中不同程度取代和不同位置取代的分子相互作用和链包装。取代包括在纤维素的每个无水葡萄糖单元中用1 - 3个羟基取代甲氧基。所提出的计算工作将回答实验员提出的关于甲基纤维素溶液热可逆凝胶化过程中形成的纤维网络内链包装的基本问题。关于甲基纤维素链如何包装和组装成具有均匀直径的原纤维，与甲基纤维素的分子量和浓度无关，仍然存在争议。最近使用小角和广角散射和显微镜进行的结构表征表明，以前的计算研究可能错误地预测了原纤维内甲基纤维素链的排列。这可能是因为过去对甲基纤维素溶液的计算研究要么使用CG模型，缺乏链的几何形状、手性和/或定向氢键相互作用，要么使用原子模型，无法捕捉到实验相关的链组装成原纤维的长度和时间尺度。因此，需要一种更好的CG模型来表示具有基本单体水平化学细节的甲基纤维素链，并使模拟能够预测链如何相互作用并在实验相关条件下包装成原纤维。建议的工作包括三个具体目标：1)利用PI最近在CG多糖模型开发方面的成功，开发一种新的甲基纤维素CG模型；2)开发一种涉及人工神经网络增强遗传算法和分子重建的计算方法，根据广泛发表的实验研究获得的实验散射结果，对原纤维内的链包装进行逆向工程；3)应用开发的方法研究广泛的甲基纤维素溶液和CG模型的扩展，用于其他纤维素衍生物的潜在研究。与之前对甲基纤维素溶液的计算研究不同，所提出的CG模型在开始时不会假设任何甲基纤维素纤维结构或链排列，例如环面链构象和环面堆积形成原纤维。相反，这种对原纤维内链如何包装的理解将从“自下而上”的组装中获得，使用由原子信息指导的短甲基纤维素链的CG模型和由长甲基纤维素链组装形成的渗透原纤维的已发表的实验散射曲线的“自上而下”分子重建。机器学习增强的CREASE（计算逆向工程分析散射实验）方法可以克服用可能不准确/不正确的分析模型拟合散射剖面的限制，并提供超出正确分析模型拟合所能提供的微观包装信息。与美国国家标准与技术研究所的散射专家合作，将促进这种机器学习增强的CREASE在甲基纤维素溶液以外材料的更广泛使用和测试。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。

NONTECHNICAL SUMMARYCellulose is an abundant naturally found biopolymer and methylcellulose is obtained via non-toxic chemical substitution of cellulose. The chemical process that leads to formation of methylcellulose from cellulose disrupts the molecular-level interactions that cause the insolubility of cellulose in water. The resulting improved solubility of methylcellulose in water and its abundant natural raw material (cellulose) makes aqueous solutions of methylcellulose useful in many applications as food additives, paint removal agents, adhesives, emulsifying agents, and biodegradable packaging materials. To tailor the physical properties of aqueous solutions of methylcellulose for use in the above applications, there is a need for fundamental research to understand the physical properties of aqueous solutions and gels of methylcellulose as a function of temperature and concentration, and the underlying molecular structure of methylcellulose chains that drive these properties. Molecular simulations are cheap and effective in comparison with real experiments and serve as valuable microscopic tools providing molecular insight into structure within polymer solutions. For methylcellulose solutions and gels, however, there are only a handful of computational studies partly due to the complexity of these materials and partly due to lack of good molecular models. This project is aimed at developing better methylcellulose models and computational methods to enable fundamental studies of structure of methylcellulose chains in water at different temperatures and concentration and guide the practical use of methylcellulose solutions in a variety of day-to-day applications. The PI will integrate the model and computational method development into her interdisciplinary molecular modeling and simulation of soft materials elective course at University of Delaware. This course is offered once every two years and is open to both undergraduates and graduate students in the (Chemical and Materials) engineering and physical sciences (Physics, Chemistry) programs. The PI will also include the data science aspects of the project in the Chemical Engineering undergraduate course on Probability and Statistics for Chemical Engineers. To improve recruitment and retention of women scientists within the computational materials field, PI Jayaraman will continue to organize seminars/talks like the successful WELCOME: Women ExceLling in COmputational Molecular Engineering virtual monthly seminar series that she initiated in 2020-21. Such virtual seminars will continue to provide networking opportunities to women graduate students and early career researchers within the research community and help with recruitment and retention of women and URM researchers in STEM careers. TECHNICAL SUMMARYThe PI proposes to develop new coarse-grained (CG) models and computational approaches to understand the molecular interactions and chain packing in aqueous solutions of methylcellulose for varying degrees of substitution and varying placement of these substitutions. Substitutions involve replacing a methoxy by one-to-three hydroxyls in each anhydroglucose unit of cellulose. The proposed computational work will answer fundamental questions raised by experimentalists regarding chain packing within fibrillar networks formed during thermoreversible gelation of methylcellulose solutions. There is still debate over how methylcellulose chains pack and assemble into fibrils with uniform diameters independent of methylcellulose molecular weight and concentration. Recent structural characterization using small- and wide-angle scattering and microscopy suggest that previous computational studies may have predicted methylcellulose chain packing within the fibrils incorrectly. This could be because past computational studies on methylcellulose solutions have either used CG models that lack chain geometry, chirality and/or directional hydrogen bonding interactions or used atomistic models which cannot capture the experimentally relevant length and time scales of chain assembly into fibrils. Thus, there is a need for a better CG model to represent methylcellulose chains with essential monomer-level chemical details and enable simulations to predict how chains interact and pack into fibrils at experimentally relevant conditions. The proposed work consists of three specific aims: 1) develop a new CG model for methylcellulose, leveraging recent success with CG polysaccharide model development by the PI, 2) develop a computational approach involving an artificial neural network enhanced genetic algorithm and molecular reconstruction to reverse engineer the chain packing within fibrils from experimental scattering results obtained from extensive published experimental studies, and 3) apply the developed approaches to study a broad range of methylcellulose solutions and extension of CG model for potential studies of other cellulose derivatives. Unlike previous computational studies of methylcellulose solutions, the proposed CG model would not assume at the start any methylcellulose fibril structure or chain packing, such as toroidal chain conformations and stacking of toroids to form fibrils. Instead, this understanding of how chains pack within fibrils will be obtained from ‘bottom up’ assembly using the CG model of short methylcellulose chains guided by atomistic information and ‘top down’ molecular reconstruction of published experimental scattering profiles of percolated fibrils formed from assembly of long methylcellulose chains. The machine learning enhanced CREASE (computational reverse engineering analysis for scattering experiments) approach may overcome the limitation of fitting scattering profiles with a possibly inaccurate/incorrect analytical model and provides microscopic packing information beyond what a correct analytical model fit would provide. A wider use and testing of this machine learning enhanced CREASE for materials beyond methylcellulose solutions will be facilitated by a collaboration with scattering experts at the National Institute of Standards and Technology.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.