Stabilizing therapeutic protein solutions: Optimisation and Evaluation of Excipient Properties using MD, QSAR and Synthesis

稳定治疗性蛋白质溶液：使用 MD、QSAR 和合成优化和评估赋形剂特性

• 批准号: 2283681
• 金额: --
• 依托单位: University of Nottingham
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2283681
• 关键词: Stabilizing; therapeutic; protein; solutions; Optimisation

Aggregation of therapeutic proteins including antibodies has been identified as a major challenge to their commercialisation and clinical use. Aggregation can cause reduced biological activity, increased viscosity and potentially enhanced immunogenicity. These issues have resulted in the employment of several types of excipients in current therapeutic protein formulations (Humira (adalimumab) contains 0.1% W/V polysorbate 80 (TWEEN80), Raptiva (efalizumab) 0.2% W/V polysorbate 20 (TWEEN20), Orencia (abatacept) contains poloxamer 188 (pluronic F-68). These non- ionic surfactants have been chosen mainly because they have well established safety profiles, rather than outstanding performance as protein stabilizing agents.The proposed research will have two main activities. Using existing excipient structures with known properties and literature data, a combination of in silico molecular dynamics and where sufficient related excipient structures have been studied, in silico QSAR studies incorporating machine learning will be undertaken.Using techniques we have established through previous CDT projects (see Mackenzie, JCTC 2015, 11, 2705-2713) molecular dynamics simulations will study the interactions between selected excipients and proteins with established 3-dimensional structure. These will characterise the locations, strengths, and (possibly) timescales of interactions, and the effects that excipient interactions haver on the structure of the protein.Machine learning methods will be used to model the relationships between chemical structures of the excipients and their protein binding affinities. A variety of 2D- and 3D-representations will be used to describe the chemical structures. Graph-based approaches capture the connectivity of different atom types and are quick to compute and readily generalizable. Interaction fields are derived from the 3D structures of the molecules and can be extended to incorporate conformational sampling. These representations will be used to train machine learning methods, including support vector machines, neural networks and random forests. New experimental data will be used to refine the machine learning methods, increasing their predictive ability. Conversely, the models will be used to guide subsequent experiments, in order to test specific hypotheses about the importance of various physicochemical properties and to identify more effective excipients.The results of these studies will inform and guide the development of new protein stabilization excipients, both through moderate modifications such as homologation/monomer extension of existing surfactants and more disruptive changes such as the inclusion of new functional groups that can change the LogP/LogD; rotational freedom (i.e addition of E/Z alkenes or cyclopropyl/diol groups to unsaturated ); H-bonding ability; -stacking ability; or inclusion of charged groups such as the guanidine group as found in arginine (an excipient that can ion-pair, H-bond with carboxylate groups and form -cation interactions with aromatic groups). The ability of both existing and new compounds to stabilize a range of therapeutic proteins (insulin, abatacept, human serum albumin, adalimumab) in solution will then be studied using a manifold of biophysical techniques (CD, ITC, SEC, DLS, AUC) in order to determine which has the largest stabilizing effect, and to quantify the surfactant structure and activity.The student will therefore be trained in a range of complementary techniques including computational methods, organic synthesis and compound characterization and a range of biophysical techniques for characterizing protein-excipient mixtures. This project fits within the 21st Century Products priority, Healthcare Technologies (developing future therapies) and manufacturing for the future themes of the EPSRC.Project aligned to Predictive Pharmaceutical Sciences, Advanced Product Design and Complex Product Characterisation

包括抗体在内的治疗性蛋白质的聚集已被确定为其商业化和临床应用的主要挑战。聚集可导致生物活性降低、粘度增加和潜在的免疫原性增强。这些问题导致在目前的治疗性蛋白质制剂中使用几种类型的赋形剂（Humira（阿达木单抗）含有0.1% W/V聚山梨酯80（TWEEN 80），Raptiva（依法利珠单抗）含有0.2% W/V聚山梨酯20（TWEEN 20），Orencia（阿巴西普）含有泊洛沙姆188（普朗尼克F-68）。选择这些非离子表面活性剂主要是因为它们具有良好的安全特性，而不是作为蛋白质稳定剂的出色性能。使用具有已知性质的现有辅料结构和文献数据，结合计算机分子动力学和已研究的足够相关辅料结构，将进行结合机器学习的计算机QSAR研究。（参见麦肯齐，JCTC 2015，11，2705-2713）分子动力学模拟将研究所选赋形剂与具有已建立的三维结构的蛋白质之间的相互作用。这些将描述相互作用的位置、强度和（可能的）时间尺度，以及赋形剂相互作用对蛋白质结构的影响。机器学习方法将用于建模赋形剂的化学结构与其蛋白质结合亲和力之间的关系。将使用各种2D和3D表示来描述化学结构。基于图的方法捕获不同原子类型的连接性，并且计算速度快，易于推广。相互作用场来自分子的3D结构，并且可以扩展到包含构象采样。这些表示将用于训练机器学习方法，包括支持向量机，神经网络和随机森林。新的实验数据将用于改进机器学习方法，提高其预测能力。相反，这些模型将用于指导后续实验，以检验有关各种理化性质重要性的特定假设，并确定更有效的辅料。这些研究的结果将为新的蛋白质稳定辅料的开发提供信息和指导，通过适度的修饰，如同源化/现有表面活性剂的单体扩展和更多的破坏性变化，例如包含可以改变LogP/LogD的新官能团;旋转自由度（即E/Z烯烃或环丙基/二醇基团加成至不饱和的）;或者包括带电基团，例如在精氨酸中发现的胍基（一种赋形剂，其可以与羧酸根基团离子配对、氢键结合并与芳族基团形成阳离子相互作用）。现有和新化合物稳定一系列治疗性蛋白质的能力（胰岛素、阿巴西普、人血清白蛋白、阿达木单抗）在溶液中的浓度进行研究（CD、ITC、SEC、DLS、AUC）以确定哪一种具有最大的稳定作用，并量化表面活性剂的结构和活性。因此，学生将接受一系列补充技术的培训，包括计算方法，有机合成和化合物表征以及一系列用于表征蛋白质-赋形剂混合物的生物物理技术。该项目符合21世纪世纪产品优先级、医疗保健技术（开发未来疗法）和EPSRC未来主题的制造。该项目与预测药物科学、先进产品设计和复杂产品表征相一致