GAIUS - Maintenance activities for the sustainability of AUGUSTUS

AUGUSTUS是基因组序列中基因结构注释的工具。结构基因组注释是在基因组中寻找基因并鉴定其外显子-内含子结构的经典生物信息学任务。它通常借助于RNA-Seq和相关基因组的同源性数据。全面的最新基因组注释还需要一个具有数千个参数的进化枝特定统计模型来调整目标基因组。基因预测任务频繁地在大多数新测序的基因组上进行。在最近对基因组注释方法的独立评估中，AUGUSTUS属于最准确的程序，例如，它对人类蛋白质编码基因的最高基因水平敏感性达到61%。多年来，我们一直观察到引用数量的上升，以及非常高的下载和web服务提交数量。AUGUSTUS的开发是在不同的研究项目中进行的，例如同源整合的新方法，自动训练，RNA-Seq整合，替代剪接或多基因组注释。目前，AUGUSTUS作为用c++编写的开源代码和通过两个web服务提供给用户。然而，AUGUSTUS的可用性和代码质量还没有得到解决。缺乏可用性使得不太懂it的生物学家使用AUGUSTUS尤其困难。当前的可用性缺陷花费了许多用户宝贵的时间，并迫使其他用户选择更方便但不太合适的选择，这可能最终导致错误的结论或以后失败的实验。此外，目前还没有一个系统可以在研究相关物种的不同研究小组之间共享特定物种的参数。缺乏对代码质量的关注使得其他研究人员很难做出贡献。此外，由于源代码开发目前发生在Greifswald大学的内部存储库中，并且没有适当的问题管理来处理更改请求和错误报告，因此其他研究人员很难参与AUGUSTUS的开发。因此，我们建议在GAIUS项目中解决上述问题。我们将通过更好的文档、开发更简单的接口和统一的管道脚本来提高可用性。参数和数据共享的存储库将允许用户直接受益于其他用户在相关基因组上的工作。这也将支持可复制的研究。通过Debian软件包和管道虚拟化，AUGUSTUS的本地安装将变得更加简单。源代码和问题管理将被解决，例如通过使用GitHub。WebAUGUSTUS部署基础设施将通过采用DevOps方法来改进，以促进未来的更新；非常重要的是，通过对现有代码进行额外的测试和重构，源代码的技术深度将会降低。

AUGUSTUS is a tool for the structural annotation of genes in genomic sequences. Structural genome annotation is the classical bioinformatics task of finding genes and identifying their exon-intron structure in a genome. It is commonly aided by RNA-Seq and by homology data from related genomes. Comprehensive state of the art genome annotation also requires that a clade-specific statistical model with thousands of parameters is adjusted to the target genome. The task of gene prediction is carried out frequently and on most newly sequenced genomes. In the most recent independent assessment of genome annotation methods, AUGUSTUS belonged to the most accurate programs, for example, it achieved with 61% the highest gene-level sensitivity on human protein-coding genes. For years, we have been observing a rising number of citations as well as a very high number of downloads and web service submissions.The development of AUGUSTUS was performed in different research projects, e.g. on new methods for homology integration, for automatic training, for RNA-Seq integration, alternative splicing or multi-genome annotation. Currently, AUGUSTUS is available to users as open source code written in C++ and through two web services. However, the usability of AUGUSTUS and code quality were not yet addressed. The lack of usability makes it especially difficult for less IT literate biologists to use AUGUSTUS. Current usability deficiencies cost many users valuable time and push other users towards choosing a more convenient but a less suited choice, that may ultimately result in wrong conclusions or failed experiments at a later time. Furthermore, there is currently no system in place to share species-specific parameters across different research groups that study related species. The lack of focus on the code quality makes it difficult for other researchers to contribute. Moreover, since the source code development currently happens in internal repositories at the University of Greifswald and there is no issue management in place to handle change requests and bug reports, it is difficult for other researchers to get involved in the development of AUGUSTUS.We therefore propose to address the above mentioned issues in the GAIUS project. We will improve usability through better documentation, development of easier interfaces and a unified pipeline script. A repository for parameter and data sharing will allow users to benefit directly of other users' work on related genomes. This will also support replicable research. The local installation of AUGUSTUS will become simpler via a Debian package and pipeline virtualization. Source code and issue management will be addressed e.g. by using GitHub. The WebAUGUSTUS deployment infrastructure will be improved through the adoption of DevOps methods to facilitate future updates; and very importantly, the technical depth of source code will be reduced through additional tests and refactoring of existing code.