Development and Maintenance of RepeatMasker and RepeatModeler

项目摘要 哺乳动物和大多数其他真核生物基因组含有大量的散布重复序列（IR），大多数 其是不同衰变水平的转座因子（TE）的拷贝。他们的存在使许多人 基因组序列分析，但他们在早期分析阶段的准确识别可以减少这些 并发症除了他们的普遍性，在过去的几十年里，研究界已经成为 他们对基因组活动和进化的巨大影响非常熟悉。 每一个物种都暴露在一套独特的、复杂的TE中， 从三亿年前到现在这些TE被发现和重建的de novo发现方法，通常由我们的RepeatModeler工具，而它们的副本则由我们的 RepeatMasker软件。从头方法可以以合理的速度创建TE库，但产品远远不够。 从所需的质量，可以达到手工策展。随着最近的爆炸性增长， 物种，这些完成步骤，也许永远不会完全自动化，现在形成了一个严重的瓶颈，基因组 由于缺乏人力和专门知识，分析结果，特别是由不同的专家提出的结果， 不同研究小组的方法缺乏一致性，而且存在冗余。而且 已经创建了高质量文库的基因组的注释没有跟上文库的发展。 由于重新分析的计算负担而导致的改进。 在这个提议中，我们描述了一个通过推广和改进TE对齐来重构RepeatMasker的计划 裁定，切换到以家庭为中心的搜索策略，支持增量重新分析，改善 注释报告和支持集群环境。针对改进方法的需要， 我们建议对RepeatModeler的核心发现进行重大更改 算法，开发了一种新的模型扩展工具，并。此外，我们将扩展我们的新方法， 利用多物种比对和祖先重建，并利用它们建立一个全面的 哺乳动物TE文库。

Project Summary Mammalian and most other eukaryotic genomes contain a large number of interspersed repeats (IRs), most of which are copies of transposable elements (TEs) at varying levels of decay. Their presence complicates many genome sequence analyses, but their accurate identification in an early analysis stage can reduce these complications. In addition to their pervasiveness, over the last decades the research community has become widely familiar with their enormous impact on genome activity and evolution. Every species has been exposed to a unique, complex set of TEs leaving recognizable copies from as long ago as 300 million years to as recently as the present day. These TEs are uncovered and reconstructed by de novo discovery methods, often by our RepeatModeler tool, while their copies are then annotated by our RepeatMasker software. De novo methods can create TE libraries at a reasonable pace, but the product is far from the desired quality that can be reached by hand curation. With the recent explosive growth in sequenced species, these finishing steps, perhaps never fully automatable, now form a severe bottleneck in genome analyses due to a lack of manpower and expertise, while the results, especially when produced by different methods from different research groups, lack consistency and suffer from redundancy. Furthermore, the annotation of genomes for which high-quality libraries have been created is not keeping up with library improvements due to the computational burden of re-analysis. In this proposal, we describe a plan to refactor RepeatMasker by generalizing and improving TE alignment adjudication, switching to a family-centric search strategy with support for incremental re-analysis, improving annotation reporting and supporting cluster environments. Responding to the need for improved methods for automated TE library generation we propose making significant changes to RepeatModeler’s core discovery algorithms, develop a novel model extension tool, and. In addition, we will extend our novel methods for exploiting multi-species alignments and ancestral reconstructions and utilize them to build a comprehensive mammalian TE library.