Comparative Genomics to Identify Functional Blocks & HGT

随着越来越多物种的基因组被测序，很明显， 确定人类基因组中区域功能的强大技术是通过与 其他物种。这种理解对疾病诊断和专门药物和疫苗的影响 设计是明确的。类似地，细菌之间的基因比较可以揭示功能上 在传染病的发展中很重要，并再次帮助药物和疫苗的设计。该项目有两个 主要研究目标。一个是开发方法学，用于寻找功能性预测特征， 非编码序列（NCS）在多个物种中高度保守，另一个是开发新的 水平基因转移（Horizontal Gene Transfer，HOT）。基因组的比较是 为了实现他们的第一个目标，研究人员计划整合基因组序列数据，这些数据由 他们的合作者，实验和文献数据，如微阵列表达数据，GO功能， 附近基因的注释和ChIP-芯片数据。结果将用于评价功能相关性，如果 每个NCS的任何一个，然后根据可测量的协变量定义预测功能的特征， 序列结构例如，如果序列特征表征NCS，其最近的基因贡献于 那么接近具有相同特征的NCS的未知基因将是主要候选者 用于询问该功能。研究人员建议通过以下方法解决这个问题：1）。开发非标 基于监督学习算法的聚类方法的类型，例如，随机森林，2）表示 NCS的随机模型的参数，并确定适当的阈值，模型拟合， 其他Monte Carlo方法。 在第二个主题下，研究人员提出了两种不同的方法来确定是否 在细菌中发生了功能显著的HGT。第一种方法是采用一个已知的功能重要的 家庭（NIFgenes），其中HGT是一个有争议的问题，并制定定量措施，他们希望将 得出一个明确的结论。他们打算完善不同物种基因之间的相似性度量，例如 BLAST评分，校正进化距离，他们将计算这些措施对NIF基因， 不同的物种，对基因已知的HGT（抗生素免疫赋予基因）和基因非常 不太可能是HGT（核糖体蛋白）。第二种方法是寻找16岁以下的人 RNA在细菌物种的大量亚群中是保守的，否则它们只是远亲。 数学和统计挑战包括：在方法I下，将基因与 不同的突变率;为HGT与非HGT设计适当的分类器，并计算适当的 当基因不是HGT时将其分类为HGT的概率的估计，反之亦然;在方法II下， - 通过考虑系统发生树拓扑来扩展现有的用于检测大内含物的方法， 橙色的长度。

As the genomes of more and more species are sequenced it has become apparent that one of the mosl powerful techniques for determining region function in the human genome is by comparison to the genomes of other species. The implications of such understanding for disease diagnosis and specialized drug and vaccine design are clear. Similarly, genpmic comparison between bacteria can reveal regions which are functionally important in the development of infectious diseases and again aid drug and vaccine design. This project has two primary research goals. One is the development of methodology for finding functionally predictive signatures of non-coding sequences (NCS)highly conserved across multiple species, and the other is to develop novel approaches for detecting Horizontal Gene Transfer (HOT). The comparison of genomes is the common thread in this research.In pursuit of their first goal, the investigators plan to integrate genomic sequence data, provided by their collaborators, with experimental and literature data, such as microarray-expression data, GO-functional- annotation for nearby genes, and ChlP-Chip data. The results will be used to evaluate the functional relevance, if any, of each NCS and then to define a signature predictive of function in terms of measurable covariates and sequence structure. For instance, if a sequence signature characterizes NCS whose nearest genes contribute to a particular function then an unknown gene close to an NCS with the same signature would be a prime candidate for interrogation of that function. The investigators propose to attack this problem by 1). Developing non standard types of clustering methods based on supervised learning algorithms, e.g.,Random Forests, 2) Representing the NCS by the parameters of a stochastic model and determining appropriate thresholds for model fitting by using resampling and other Monte Carlo methods. Under the second topic, the investigators propose two different approaches for determining whether Functionally significant HGT has occurred in bacteria. The first approach is to take a known functionally important family(NIFgenes) for which HGT is a matter of dispute, and devise quantitative measures which they expect will enable a firm conclusion. They intend to refine similarity measures between genes in different species ,such as BLAST scores, corrected for evolutionary distance They will compute these measures for pairs of NIF genes in different species , pairs of genes known to be HGT (antibiotic immunity conferring genes) and genes very unlikely to be HGT (ribosomal proteins). The second approach is to look for anomalously long stretches of 16s RNA conserved within substantial subsets of bacterial species which are otherwise only distantly related. Mathematical and statistical challenge include: Under approach I, standardizing comparisons of genes with different mutation rates; devising an appropriate classifier for HGT vs. non HGT, and computing appropriate estimates of the probability of classifying a gene as HGT when it isn't and vice versa; Under approach II, extending existing methods for detecting large inclusions by taking into account phylogenetic tree topology and oranch lengths.