A step change in the modelling of enzymatic catalysis

对酶过程的深入理解需要原子水平的描述，这往往超出了实验技术的范围，但可以通过与计算化学的协同作用来实现。这个纯计算项目提出了一种改变酶活性部位模型的方法，它建立在一种名为FFLUX的新型力场方法的基础上[1]。这种方法是基于对分子系统内原子的最佳现代定义[2]。FFLUX使用机器学习来训练这些原子，使其在以前未见过的原子构型中“知道该做什么”。换句话说，从量子力学的角度讲，酶活性部位的原子之间的能量和作用力是很容易获得的。因此，无论是静态的还是动态的，都可以快速、准确地探索活性中心的势能面。在酶工程中使用FFLUX的一个最合适的例子是萜类化合物。它们是最丰富和最大的一类(75,000英镑)天然产品[3]。大多数常见于植物中，具有从物种间交流到细胞内信号传递和防御捕食性物种的生物学作用。它们的商业用途广泛，如制药、除草剂、调味品、香料和生物燃料。分子动力学模拟表明，单萜合成酶类的酶在反应周期中(在初始底物结合后)不会发生大规模的构象变化，这使得它们成为基于结构的蛋白质工程的有吸引力的靶标，以扩大其对替代单萜支架的催化范围。这类非常重要的化合物在MIB中得到了彻底的研究，这丰富了与实验者互动的范围。该项目的目的是在模拟酶的活性部位的现实主义上创造一个阶段性的变化。对反应进行严格但容易理解的量子力学描述对于最终的进展至关重要。只有使用FFLUX对酶的反应机理有了详细的原子洞察，才能正确地使未来酶的设计合理化。

A deep understanding of enzymatic processes requires an atomic-level description, which is often beyond the reach of experimental techniques but can be achieved through synergy with computational chemistry. This purely computational project proposes a step change in the modelling of an enzyme's active site by building on the foundations laid by a novel force field method called FFLUX[1]. This method is based on the best modern definition of an atom inside a molecular system[2]. FFLUX uses machine learning to train these atoms to "know what to do" in a previously unseen atomic configuration. In other words, quantum mechanically accurate energies and forces between atoms in an enzyme's active site are readily available. Hence, the potential energy surface of the active site can be quickly and accurately explored, both statically and dynamically. A most suitable prototype case for using FFLUX in enzyme engineering is that of terpenoids. They are the most abundant and largest class (>75,000) of natural products[3]. Most are commonly found in plants, with biological roles ranging from interspecies communication to intracellular signalling and defence against predatory species. Their commercial use is wide ranging as pharmaceuticals, herbicides, flavourings, fragrances and biofuels. Molecular dynamics simulations suggest that the monoterpene synthase class of enzymes do not undergo large-scale conformational changes during the reaction cycle (after initial substrate binding), making them attractive targets for structured-based protein engineering to expand their catalytic scope toward alternative monoterpene scaffolds. This very important class of compounds has been thoroughly studied in the MIB, which enriches the scope for interaction with experimentalists. The aim of the project is to create a step change in the realism of modelling an enzyme's active site. A rigorous yet accessible quantum mechanical description of a reaction is vital for ultimate progress. Only with detailed atomic insight into the mechanism of an enzymatic reaction, using FFLUX-for-enzymes, can one correctly rationalise the design of future enzymes.