PIT-DB: A Resource for Sharing, Annotating and Analysing Translated Genomic Elements

2001年人类基因组的发表被正确地誉为一项重大的科学成就，但十多年后，我们仍然远未完全了解基因组的结构和其中各种元素的作用。虽然基因组的蛋白质编码区在被测序后不久就被识别并用于注释基因组，但更多的外来基因组元件随后引起了人们的兴趣，包括伪基因、非编码RNA和短开放阅读框架(SORF)。近年来，后基因组生物分析技术，如RNA-seq转录组学(它告诉我们哪些基因组元件是表达的)和基于质谱学的蛋白质组学(它告诉我们哪些表达的元件被翻译成肽或蛋白质)帮助我们在基础水平上加深了对人类基因组的理解。就在今年，发表在《自然》杂志上的两项蛋白质组学研究引起了轰动，因为它们表明，没有实验证据表明，几种被广泛接受的基因组元素可以表达蛋白质，而基因组中其他以前被认为不是蛋白质编码的区域实际上可以产生蛋白质。如果这是深入研究的人类基因组的情况，我们必须假设对较少研究的物种(所谓的非模式生物)的基因组注释甚至更不准确。我们最近开发(并测试并发表)了一种名为蛋白质组学由转录组学信息(PIT)的方法，该方法快速生成大量的基因组注释，并得到多个实验证据来源的支持。在PIT中，每个样本都使用RNA-seq和蛋白质组质谱进行分析，这两种分析的数据结合在一起，提供了一份观察到的蛋白质和任何其他翻译的基因组元素(TGE)的清单，以及支持这些观察的详细转录和光谱证据。与传统蛋白质组学相比，PIT的优点在于不需要先验序列知识，因此可以检测到新的TGE(无论是蛋白质还是其他更奇异的特征)。RNA-SEQ本身可以用来快速生成基因组注释，而不需要事先知道，但是如果没有PIT的质谱学步骤，对这些注释的信心是有限的，并且不能保证转录的元件确实被翻译。在最近的BBSRC TRDF项目中，我们开发了易于使用的基于Web的软件工作流，在流行的Galaxy平台上实现，以可重复的方式处理来自PIT实验的数据和统一格式的输出文件。事实证明，这对回答个别的生物学问题非常有用，但目前还没有有意义的方式来分享PIT实验的结果。在这个项目中，我们建议通过开发PIT-DB来填补这一空白，这是一个可通过网络访问的关于PIT产生的结果的数据库。这个公开的数据库将立即填充布里斯托尔大学对各种物种进行的实验数据，但将积极鼓励其他小组提交自己的数据。将多个坑实验的数据保存在一个数据库中将带来令人兴奋的新科学见解。除了简单地允许研究人员分享他们在各个PIT实验中的结果外，PIT-DB还将汇集来自多个实验的单个新型TGE的信息，以便为每个TGE积累证据。通过使用重复实验的数据来提高结果的质量是科学中的一个基本概念，在社区范围内这样做的效用已被其他生物信息学数据库如EnSembl、UniProt和Pride反复证明。除了个别感兴趣之外，PIT-DB中证据充分的TGE将为数据库中存在数据的所有物种提供大量实验得出的(而不是计算预测的)基因组注释。

The publication of the human genome in 2001 was rightly hailed as a major scientific achievement, but over a decade later we are still far from a complete understanding of the structure of the genome and the role of the various elements within it. While protein coding regions of the genome were identified and used to annotate the genome soon after it was sequenced, many more exotic genomic elements have subsequently attracted interest including pseudogenes, non-coding RNAs and short open reading frames (sORFs).In recent years, post-genomic bioanalytical techniques such as RNA-seq transcriptomics (which tells us which genomic elements are expressed) and mass spectrometry based proteomics (which tells us which of the expressed elements are translated into peptides or proteins) have helped refine our understanding of the human genome at a fundamental level. Just this year, two proteomics studies published in Nature caused a stir by showing that no experimental evidence could be found for the expression of several genomic elements widely accepted to code for protein, while other regions of the genome that were not previously thought to be protein coding were in fact found to produce proteins. If this is the situation for the intensively studied human genome, we must assume that the genome annotations for less studied species (so called non-model organisms) are even less accurate.We recently developed (and tested, and published) a methodology called proteomics informed by transcriptomics (PIT) that rapidly generates large numbers of genome annotations underpinned by multiple sources of experimental evidence. In PIT, every sample is analysed using both RNA-seq and proteomic mass spectrometry and the data from these two analyses integrated to provide a list of observed proteins and any other translated genomic elements (TGEs), together with the detailed transcriptomic and spectral evidence that underpins these observations. The beauty of PIT compared with traditional proteomics is that no prior sequence knowledge is needed, so novel TGEs (be they proteins or other more exotic features) can be detected. RNA-seq can be used by itself to rapidly generate genome annotations without prior knowledge, but without PIT's mass spectrometry step the confidence in these annotations is limited and there is no guarantee that transcribed elements actually get translated.In a recent BBSRC TRDF project we developed easy to use web-based software workflows, implement in the popular Galaxy platform, to process the data from PIT experiments in a repeatable way with uniformly formatted output files. This has proven very useful for answering individual biological questions, but there is currently no meaningful way to share the results of PIT experiments. In this project we propose to plug this gap by developing PIT-DB, a web-accessible database of results produced by PIT. This publicly available database will immediately be populated with data from experiments conducted on various species at the University of Bristol, but other groups will be actively encouraged to submission their own data.Having data from multiple PIT experiments in one database will deliver exciting new scientific insights. As well as simply allowing researchers to share their results from individual PIT experiments, PIT-DB will pool information about individual novel TGEs from multiple experiments so evidence can be accumulated for each individual TGE. Improving the quality of results by using data from replicate experiments is a fundamental concept in science and the utility of doing this on a community-wide basis has been repeatedly demonstrated by other bioinformatics databases such as Ensembl, UniProt and PRIDE. As well as being of interest individually, the well evidenced TGEs in PIT-DB will provide large numbers of experimentally derived (as opposed to computationally predicted) genome annotations for all of the species for which data is present in the database.