Engineering enzymes in all their dimensions: from inhibition to biotransformation

我们研究的首要目标是通过将更多基于工程知识的设计和创造纳入酶工程，加速应用蛋白质的实验室进化。酶的应用范围从药物合成的立体选择性反应到增值产品和功能化材料的大规模转化，酶以全球可持续的方式加速化学转化。然而，需要对酶进行工程改造，以拓宽它们催化的反应范围;我们还需要知识来抑制药理学相关的酶。更好的酶及其抑制剂的设计越来越多地通过计算方法获得信息，这些方法将假设的设计排列成最有可能正确的设计，从而集中实验努力。 机器学习将大大提高计算预测的能力。机器学习的预测准确性源于为学习提供的信息的质量和数量。目前关于酶结构/功能关系的知识状态还不能可靠地为酶工程的预测算法提供燃料。我们在即将到来的资助期内的具体重点是将酶功能的分子细节转化为机器学习算法的描述符。描述符各不相同：我们将通过执行计算模拟、创建突变酶库、功能筛选、结构表征和验证的循环，研究酶活性位点和底物进入/退出路径的多种物理和化学特征。我们将通过工程化属于不同类别的酶来测试描述符的预测能力，因为我们努力理解酶工程的一般规则，而不是在单一类型的酶上过度训练。 使我们的酶工程方法与众不同的一个重要创新是考虑酶的运动。它们的运动范围从快速的侧链波动到缓慢的域运动。这第四个维度是酶功能的内在描述符，并且可以作为多个突变之间表现出的协同作用的基础，特别是当它们集中在结合和活性位点区域内时。使用实验和计算方法，我们将更好地了解特定类型的蛋白质运动对酶催化的影响，以及运动如何发挥酶进化的潜力。 我们的产出将包括针对新出现的抗生素耐药性来源的更强大的抑制剂，以及用于高效生产增值产品的生物催化剂。重要的是，所获得的知识将用于计算，以增强计算酶工程的预测能力。

The overarching goal our research is to accelerate the laboratory evolution of applied proteins, by putting more engineering' knowledge-based design and creation into enzyme engineering. With applications ranging from stereoselective reactions for pharma synthesis to large-scale conversion for value-added products and functionalized materials, enzymes accelerate chemical transformations in a globally sustainable way. However, enzymes need to be engineered to broaden the spectrum of reactions they catalyze; we also need knowledge to inhibit enzymes of pharmacological relevance. The design of better enzymes and of their inhibitors is increasingly informed by computational methods that rank hypothetical designs into those most likely to be correct, thereby focussing experimental efforts. Machine learning will dramatically increase the power of computational predictions. The predictive accuracy of machine learning stems from the quality and quantity of the information provided for learning' to occur. The current state of knowledge on the enzyme structure/function relationship does not yet reliably fuel predictive algorithms for enzyme engineering. Our specific focus in the upcoming funding period is to translate molecular details of enzyme function into descriptors for machine-learning algorithms. The descriptors are varied: we will investigate multiple physical and chemical features of enzyme active-sites and of substrate entry/exit paths, by performing cycles of computational simulations, creation of mutated enzyme libraries, functional screening, structural characterization and validation. We will test the predictive capacity of the descriptors by engineering enzymes belonging to different classes because we strive to understand general rules of enzyme engineering rather than overtraining' on single type of enzyme. An important innovation that distinguishes our approach to enzyme engineering is to consider the motions of enzymes. Their motions range from rapid side-chain fluctuations to slow domain motions. This fourth dimension is an intrinsic descriptor of enzyme function and can underlie the synergy exhibited between multiple mutations, particularly when they are focused inside the binding and active-site area. Using experimental and computational methods, we will gain a better understanding of the impact of specific types of protein motions on enzyme catalysis, and how motions play into potential for an enzyme to evolve. Our output will include more powerful inhibitors of an emerging source of antibiotic resistance, and biocatalysts for the efficient generation of value-added products. Importantly, the knowledge gained will feed into calculations to enhance the predictive capacity of computational enzyme engineering.