LIKELIHOOD ALIGNMENT AND EVOLUTIONARY MODELS

对齐是关于进化对应关系的假设 DNA 或蛋白质序列。序列比对对于研究至关重要 分子进化。此外，它们还可用于检测高度 保守的功能域，并且是一些策略的组成部分 搜索序列数据库。广泛使用的对齐技术依赖于 未阐明且不清楚的进化假设。特设的 这些技术的基础让人对它们产生的比对产生怀疑。 拟议的研究将涉及对齐方法的开发 基于显式进化模型的推论。 特别 我们将注意使这些模型尽可能真实。 核苷酸组成的区域异质性、 进化速率和相对一般的长度分布 将允许插入删除事件。 进化模型将为 应用包括：蛋白质二级结构的推断， 进化参数的估计和确定 对齐中的可靠区域。另外，还有一种方法可以同时 将介绍比对多个序列并重建系统发育。 这项研究产生的序列分析技术将全部 具有可能性基础，并且可以利用可能性这一事实 统计推断的方法已得到深入研究。 的 这些技术的进化视角和概率本质 应该结合起来使它们比以前提出的更准确。

An alignment is a hypothesis about the evolutionary correspondence between DNA or protein sequences. Sequence alignments are crucial for the study of molecular evolution. In addition, they can be used to detect highly conserved functional domains and are a component of some strategies for searching sequence databases. Widely-used alignment techniques rely on evolutionary assumptions that are both unstated and unclear. The ad hoc basis of these techniques casts doubt on the alignments that they produce. The proposed research will involve development of methods for alignment inference that are based upon explicit evolutionary models. Special attention will be paid to making these models as realIstic as possible. Regional heterogeneity of nucleotide composition, heterogeneity of evolutionary rates, and relatively general distributions for lengths of insertion-deletion events will be allowed. The evolutionary models will provide an underlying framework for applications that include: the inference of protein secondary structure, the estimation of evolutionary parameters, and the determination of reliable regions in alignments. In addition, a method to simultaneously align multiple sequences and reconstruct phylogenies will be introduced. The sequence analysis techniques that arise from this research will all have a likelihood basis and can take advantage of the fact that likelihood approaches to statistical inference have been intensively studied. The evolutionary perspective and probabilistic nature of these techniques should combine to make them more accurate than those previously proposed.