Development of Bayesian hierarchical models of genetic diversity and adaptive evolution of micro-organisms in the extreme environments

像病毒和细菌这样的微生物能迅速适应强大的选择压力。我们主要关注病毒的进化，开发了贝叶斯层次模型和程序来检测分子进化事件，这些事件在病毒的适应进化中起着重要作用。种群遗传学理论认为，处于强选择压力下的濒危种群具有较高的序列进化速率。为了了解加速的序列进化是否导致其相应蛋白结构进化的可能性更高，我们分析了患者无症状期HIV -1的v3环。由于无法通过实验测量蛋白质结构的变化，我们开发了一个序列结构适应度指数。基于蛋白质结构数据库，该索引被描述为给定模板结构的序列的可能性。感染后可能性的下降趋势暗示了结构变化。通过扩展该指数，我们以似然比的形式建立了两种蛋白质之间的结合能力指数，将蛋白质作为一个复合体与独立对进行对比。通过分析流感病毒HA的长期趋势，我们发现结合在结合区域的抗体向宿主细胞受体的逃逸比结合在邻近结合位点的抗体向宿主细胞受体的逃逸强。最后，我们建立了一个贝叶斯层次模型来检测重组。如果在进化过程中发生重组，则现存序列在基因组区域之间具有不同的系统发育关系（拓扑结构）。我们开发了一种算法来计算一对拓扑之间的重组距离。通过对相邻区域拓扑之间的距离引入先验，可以可靠地估计重组点和重组断点的数量。我们还开发了一个程序来检测收敛进化。

Micro-organisms such as virus and bacteria adapt promptly to the strong selection pressures. Focusing mostly on viral evolution, we developed Bayesian hierarchical models and procedures to detect events of molecular evolution which play essential role on their adaptive evolution.The theory of population genetics suggests that endangered populations under strong selection pressure have higher rate of sequence evolution. To see if accelerated sequence evolution leads to higher chance of structural evolution of its corresponding protein, we analyzed the V3-loop of HIV -1 during the asymptomatic period in the patients. Since it is impossible to measure the variation of protein structure experimentally, we developed an index of sequence-structure fitness. Based on the database of protein structures, the index is described as a likelihood of a sequence given a template structure. Structural variation is implied by the declining trend of the likelihood after infection. By extending this index, we developed an index of binding ability between two proteins in the form of likelihood ratio which contrasts the proteins as a complex to independent pair. By analyzing the long term trend of influenza virus HA, we observed that the escape from the antibodies bound to the binding region to the receptor of the host cell is stronger than the escape from those bound to the region neighbor to the binding sites.Finally, we developed a Bayesian hierarchical model to detect recombination. If recombination occurs during the evolutionary history, the extant sequences have different phylogenetic relation (topology) between regions of the genomes. We developed an algorithm to calculate the recombination distance between a pair of topologies. By introducing a prior on the distances between the topologies of neighboring regions, it became possible to estimate the number of recombination and recombination break points reliably. We also developed a procedure to detect convergent evolution.