Aggregation and self-assembly in colloidal and biological systems

大自然在设计材料和设备方面一直很巧妙，这些材料和设备具有特殊的属性，并在生命物质中执行特定的任务。不仅要合成分子构件，它们还必须采用正确的构象并组装成功能性的上层结构。同时，它们必须避免干扰同一空间内发生的所有其他组织过程。活细胞中的许多任务都是由蛋白质完成的。这些分子是氨基酸链，折叠成复杂的结构，致力于其特定的功能。确定结构是揭示蛋白质功能的重要阶段，这通常是通过X射线晶体学实现的。为了获得有用的分辨率，有必要纯化蛋白质并生长几乎毫米大小的无缺陷晶体。然而，蛋白质已经进化到难以结晶，因为这种聚集将对其功能有害。事实上，像血红蛋白C这样的疾病是由不必要的结晶引起的。因此，寻找能够生长足够晶体的物理条件是一项困难而耗时的任务。影响蛋白质结晶倾向的一个重要因素是由于它们表面的不均匀性而导致的它们相互作用的方向性。关于方向性对结晶的影响知之甚少，这里提出的研究的一个主要目的是使用计算机模拟来研究这种影响。虽然计算机的能力继续快速增长，但它还远远不足以处理蛋白质结晶的原子代表。我们必须设计出体现蛋白质相互作用基本物理原理的粗粒度模型，并用复杂的工具进行分析，而不是采用这种蛮力方法。除了易于计算之外，这些模型的优势在于揭示了一般的基本原理，而不是特定情况下的答案。蛋白质经常将自己组织成离散的超结构，以完成一项任务。一个优雅但有害的例子是衣壳的自我组装，衣壳是病毒的外壳，包裹着它们的遗传物质。大约有一半的病毒大致呈球形（实际上是二十面体），由少量蛋白质的拷贝高效地构建而成。事实上，许多衣壳可以从它们分离的亚基可靠地组装起来，这是值得注意的，而且不容易详细解释。特别是，避免建造错误和形成完整外壳以利于许多部分碎片的能力知之甚少。在这里，简化的计算机模型可以帮助阐明可能的途径和自组装的基本热力学。这一知识可以激发针对组装阶段的抗病毒治疗，而不是针对感染本身。它也可以通过设计用于给药的微小容器来发挥积极的作用。生物分子的粗粒度建模源于为胶体科学开发的技术。胶体涵盖了广泛的分散的纳米级颗粒，日常的例子是不同的奶油，墨水和雾。在许多人造胶体中，可以对颗粒的性质进行精细控制，从而影响它们的集体行为。该提案中的进一步项目采用了胶体作为“设计者原子”的想法。例如，如何才能促使棒状分子在低密度下连接起来，从而制造出轻质导电材料？如果胶体凝胶和玻璃状材料是由大小和相互作用的混合物而不是单一成分组成的，会发生什么？计算机模拟在回答这些问题方面发挥着至关重要的作用。

Nature has been ingenious in devising materials and devices to have specialised properties and to perform specific tasks in living matter. Not only must molecular building blocks be synthesised, they must also adopt the correct conformations and assemble themselves into functioning superstructures. At the same time, they must avoid interfering with all the other organisational processes occurring within the same space.Many tasks in living cells are performed by proteins. These molecules are chains of amino acids that fold into intricate structures dedicated to their particular function. Determining the structure is an important stage in unravelling a protein's function, and this is most often achieved by x-ray crystallography. To obtain a useful resolution it is necessary to purify the protein and grow defect-free crystals up to almost millimetre size. However, proteins have evolved to be difficult to crystallise, since aggregation of that sort would be deleterious to their function. Indeed, diseases like that of haemoglobin C arise from unwanted crystallisation. Accordingly, searching for physical conditions where adequate crystals can be grown is a difficult and time-consuming task.An important factor affecting the tendency of proteins to crystallise is the directionality of their interactions with each other due to the non-uniformity of their surfaces. Very little is known about the influence of directionality on crystallisation, and a major aim of the research proposed here is to investigate the effects using computer simulation. Although computer power continues to increase apace, it is nowhere near sufficient to treat an atom-by-atom representation of protein crystallisation. Instead of such a brute-force approach, we must devise coarse-grained models that embody the essential physics of protein interactions, and analyse them with sophisticated tools. In addition to being computationally tractable, these models have the advantage of revealing general underlying principles rather than case-specific answers.Proteins often organise themselves into discrete superstructures in order to accomplish a task. An elegant but pernicious example is the self-assembly of capsids, the coats of viruses that encapsulate their genetic material. About half of all viruses are roughly spherical (in fact, icosahedral) in shape, and are efficiently built from copies of a small number of proteins. The fact that many capsids can assemble reliably from their isolated subunits is remarkable and not easy to explain in detail. In particular, the ability to avoid construction errors and to form complete shells in favour of many partial fragments is poorly understood. Here again, simplified computer models can assist by elucidating possible pathways and the underlying thermodynamics of self-assembly. This knowledge could inspire antiviral therapy targeted at the assembly stage, rather than at infection itself. It could also be turned to positive uses by designing tiny containers to administer drugs.The coarse-grained modelling of biological molecules springs from techniques developed for colloid science. Colloids cover a broad range of dispersed nanoscale particles and everyday examples are as diverse as cream, ink and fog. In many human-made colloids, it is possible to exert fine control over the properties of the particles, thereby influencing their collective behaviour. Further projects in this proposal take up the idea of colloids as ``designer atoms.'' For example, how can rod-like molecules be encouraged to connect at low densities to make light-weight electrically conducting materials? What happens to colloidal gels and glassy materials if they are composed of mixtures of sizes and interactions rather than a uniform component? Computer simulations have a vital role to play in answering these questions.