Defining the shape of evolutionary landscapes

只有当我们了解多样化的机制时，才有可能通过观察今天存在的基因多样性来推断过去。就基因及其产物而言，我们将这种多样化称为“进化”。我们进一步使用系统发育树来表示基因的自然历史：其中，每一片叶子都是一种观察，每一次分叉都是其叶子的假定祖先，而根是所有叶子的最后共同祖先。这项建议是关于如何找到最有可能的树的严格理解，以及从蛋白质的3D结构推断树的意义。 如果我们得到一些相关基因的氨基酸序列和一棵树的拓扑(形状)，就有可能计算出这棵树与序列数据匹配的可能性。主要的困难是在数量极大的其他拓扑中找到最有可能的拓扑。这项研究将所有可能的树木视为优化景观，必须寻找最佳解决方案。我们建议开发工具来定义、可视化和探索风景。有了这些，我们将能够理解为什么找到最有可能的树是如此困难。如果我们更好地理解这个问题，我们就可以改进解决方案，并对所有依赖树预测功能、对生物进行分类或发现新知识的生物信息学方法产生积极影响。 对蛋白质序列数据的研究已经进行了多年。树的形状往往反映了蛋白质通过遗传和复制的多样性。由这些序列编码的3D结构的进化机制还没有被很好地理解。在这个项目中，我们使用定义良好的推断序列树的方法来构建蛋白质结构的系统发育图。这些树讲述的故事将非常不同，可能涉及到进化时代的许多古代事件。这是因为结构的演化不像序列那样快。重建结构多样化的事件尤其具有挑战性，因为有理由相信，结构并不总是以树状方式演变。这使得它们的自然历史更类似于整个细菌基因组的进化。

It is possible to infer the past by observing the diversity of genes that exist today only if we understand the mechanics of diversification. In the case of genes and their products, we call this diversification "evolution". We further represent the natural history of genes using phylogenetic trees: where each leaf is an observation, each bifurcation an assumed ancestor to its leaves, and the root being the last common ancestor to all leaves. This proposal is about the rigorous understanding of how the most likely tree can be found, and about the meaning of trees inferred from the 3D structures of proteins. If we are given the amino acid sequence of a number of related genes and the topology (shape) of a tree, it is possible to compute how likely this tree fits the sequence data. The main difficulty is to find the most likely topology amongst an extremely large number of others. This research looks at all possible trees as an optimisation landscape which has to be searched for the best solution. We are proposing to develop tools to define, visualise and explore landscapes. With these, we will be able to understand why finding the most likely tree is so difficult. If we understand better the problem, we can improve on the solution and positively affect all bioinformatics methods that rely on a tree to predict functions, classify organisms, or discover new knowledge. Studying protein sequence data has been done for years. The shape of the tree often reflects the diversification of proteins through inheritance and duplication. The mechanism of evolution of the 3D structures encoded by these sequences is not as well understood. In this program, we are using the well defined methodologies of inferring sequence trees to build phylogenies of protein structures. The kind of stories told by these trees will be very different, reaching possibly to much ancient events in evolutionary time. This is because structures do not evolve as quickly as sequences. Reconstructing events of structural diversification is especially challenging since there are reasons to believe that structures do not always evolve in a tree-like fashion. This makes their natural histories more similar to the evolution of whole bacterial genomes.