Comparative Genomic Analysis Using Evidence Integration Frameworks

最近生物技术和测序方面的突破为复杂基因组序列的比较分析创造了许多新的机会。比较基因组分析需要建立数据库和工具来解决重要的生物学问题，包括阐明导致物种多样性、稳健性、适合性和分类学组织的保护模式和分歧模式。进化选择在不同的功能部位创造了不同的保守率，从而在不同的基因组区域产生了不同的比较特征。这些特征可以被计算方法用来改进对重要功能区域的检测，如蛋白质编码外显子、RNA基因、启动子区域、起始点和3‘UTR区。由于编码区的保守模式与内含子或基因间区的模式不同，原则上，比较计算分析可以通过使用参考基因组来显著改进对目标基因组中基因的计算识别。这项工作集中于对这些基本问题的系统研究，目的是产生一个可以针对不同物种进行训练的原型、模块化和自适应的基因识别系统。用于比较基因组分析的基因组序列建模和进化比对的模块化和自适应体系结构具有三个重要特征：a)允许对特定生物的跨物种分析进行全球模型训练的组成组织。B)一种新的机制来执行模型的贝叶斯适应被比较的特定的一对同源基因，以便改进解释，从而提高预测精度。C)灵活整合用于比较基因鉴定的其他证据来源，包括与蛋白质、EST和基因表达数据库的匹配。拟议的研究具有一些广泛的影响，将可从比较分析和发现新基因中提取的进化模式编目，建立一个高度可用的比较基因组分析软件原型，将免费分发给学术界或用于提高基因组区域注释准确性的软件原型；制作可用于进化分析和生物学研究的功能位点的比较数据库，并培训新一代计算生物学研究人员，以加深对生物过程和复杂计算方法的理解。

Recent breakthroughs in biotechnology and sequencing create numerous new opportunities for comparative analysis of complex genomic sequences. Comparative genomic analysis requires building databases and tools to address important biological questions including elucidating the molds of conservation and patterns of divergence that lead to species diversification, robustness, fitness, and taxonomical organization. Evolutionary selection creates variable rates of conservation on different functional sites thereby producing distinctive comparative signatures in different genomic regions. These signatures can be exploited by computational methods for an improved detection of functionally important regions such as protein-coding exons, RNA genes, promoters regions, initiation sites and 3'UTR regions. Since the pattern of conservation in coding regions is different from the pattern in intronic or intergenic regions, a comparative computational analysis can lead, in principle, to a significantly improved computational identification of genes in a target genome by using a reference genome. This work focuses on a systematic investigation of these fundamental questions with the goal of producing a prototype, modular and adaptive gene identification system that can be trained on different species. A modular and adaptive architecture for modeling genomic sequences and evolutionary alignment for comparative genomic analysis has three important features: a) A compositional organization that allows global model training for cross-species analysis of specific organisms. b) A novel mechanism to perform Bayesian adaptation of the model to the specific pair of orthologous genes being compared in order to refine the interpretation and thereby improve prediction accuracy. c) A flexible capability to integrate other sources of evidence for comparative gene identification that include matches to proteins, ESTs and gene expression databases. There are a number of broad implications of the proposed research, cataloging the evolutionary patterns that can be extracted from comparative analysis and discovery of novel genes, building a highly usable software prototype for comparative genomic analysis that will be distributed freely to the academic community or order to improve the annotation accuracy of genomic regions; Producing comparative databases of functional sites that can be used for evolutionary analysis and biological research, and training a new generation of computational biology researchers to attain a deeper understanding of biological processes as well as sophisticated computational methodologies.