Computational methods to predict stable biologics formulations

工程治疗蛋白质（生物制品）是发展最快的一类药物和疫苗。然而，获得满足其可制造性、稳定性和药理学功效的所有要求的蛋白质设计仍然具有挑战性。一个特别严峻的挑战是它们的边缘稳定性和对pH、温度、离子强度和机械搅拌的细微变化的高敏感性。分子先导物鉴定选择与疾病靶标结合最好的候选物，但不能确保药物剂型的可制造性和稳定性。随后通过蛋白质工程或配方进行的优化使用高度劳动和时间密集型的自动化高通量实验，但无法保证成功。这一借调将进行基础研究，将UCL开发的技术转化并嵌入工业环境，对生物开发管道产生直接影响。它将建立一个计算框架，以有效地预测有前途的生物制剂，根据精心挑选的配方的实验性能，从而节省大量的时间和资源。这一小组制剂（不同的辅料、pH值和离子强度）最能代表更广泛的潜在制剂的特征。计算分子动力学模拟在整个pH值和离子强度范围内，确定主要的蛋白质构象访问。模拟辅料与这些构象的结合（对接），评价其在每种溶液条件下相互作用的程度和强度。在UCL，基于抗体衍生疗法，这些工具已经测量了pH值，温度，离子强度和赋形剂如何影响表面极性，电荷，蛋白质动力学，蛋白质-赋形剂相互作用，以及赋形剂的优先排除，最终蛋白质稳定性。神经网络（NN）方法在伦敦大学学院已经显示出显着的潜力，从光谱测量天然蛋白质制剂预测蛋白质的稳定性。这种神经网络方法正在改变我们的日常生活，使我们能够从复杂的数据中提取预测。这项借调将把这些工具应用于IPSEN基于工程肉毒杆菌神经毒素（BoNT）开发的一类新的治疗蛋白。它将使用神经网络方法结合实验，模拟和对接数据，探索上述因素的相对平衡，并为改进配方建立一个强大的预测框架。这些将在IPSEN的生产工艺中进行确认。这将对BoNT衍生的蛋白质疗法产生直接的价值和影响，并解决这类令人兴奋的分子特有的问题。生成的预测框架也将在不同类别的蛋白质治疗剂中具有更广泛的实用性，以实现其快速和更成功的配方。这些方法将在出版物中介绍，并通过培训转移到工业中。

Engineered therapeutic proteins (biologics) are the fastest growing class of medicines and vaccines. However, it remains challenging to obtain protein designs that meet all requirements for their manufacturability, stability and pharmacological efficacy. A particularly acute challenge is their marginal stability and high sensitivity to subtle changes in pH, temperature, ionic strength, and mechanical agitation. Molecular lead identification selects for candidates that bind best to the disease target, but does not ensure manufacturability and stability in drug dosage forms. Subsequent optimisation by protein engineering or formulation uses highly labour- and time-intensive automated high-throughput experiments, yet with no guarantee of success.This secondment will undertake fundamental research to transform and embed techniques developed at UCL into an industrial setting, with immediate impact on biologic development pipelines. It will establish a computational framework to efficiently predict promising biologics formulations, based on experimental performance in carefully selected formulations, thus saving considerable time and resource. The small set of formulations (varying excipients, pH and ionic strength), best represent features of a much wider range of potential formulations. Computational molecular dynamics simulations across the pH and ionic strength range, identify the major protein conformations accessed. Simulated excipient binding (docking) to these conformations, evaluates the extent and strength of their interactions within each solution condition. At UCL, based on antibody-derived therapeutics, these tools have already measured how pH, temperature, ionic strength and excipients impact surface polarity, charge, protein dynamics, protein-excipient interactions, and preferential exclusion of excipients, and ultimately protein stability. Neural network (NN) methods at UCL have shown significant potential for predicting protein stability from spectroscopic measurements of native protein formulations. Such NN methods are transforming our daily life, enabling us to extract predictions from complex data. This secondment will apply these tools to a new class of therapeutic proteins developed by IPSEN, based on engineered botulinum neurotoxins (BoNTs). It will use NN approaches on the combined experimental, simulation and docking data, to explore the relative balance of the factors above and build a powerful predictive framework for improved formulations. These will be validated at IPSEN in their manufacturing processes. This will have immediate value and impact for BoNT-derived protein therapies, and address issues specific to this exciting class of molecules. The predictive framework generated will also have wider utility across different classes of protein therapeutics for their rapid and more successful formulation. The methods will be described in publications and transferred into industry through training.