Functional Annotations for the Pneumocystis Genome.

描述(由申请人提供)： 1999年资助了一个肺孢子虫基因组项目，目的是创建物理图谱、EST数据库和卡氏肺孢子虫(Pc)真菌病原体的汇编DNA序列。肺炎衣原体是与艾滋病毒相关的主要感染之一，肺炎(PCP)。Pc的基因组测序正在进行中：可用于分析的是几条连续的大段染色体以及一组非冗余的-1900个EST。为了充分利用不断增长的序列数据，现在需要开发在Pc中发现基因和对基因产物进行进一步注释的策略。在这个补充项目中，我们建议(与PAS-02-046的具体目标一致)开发和应用计算工具，用于Pc基因组的基因预测和注释，以促进寻找新的药物靶点和潜在的PCP药物候选。尽管越来越多的真菌基因组正在被测序，但针对真菌的基因预测程序还很少。由于A/T含量、内含子/外显子边界、启动子序列和基因密度等方面的特殊偏差，以及不同真菌基因组之间在组织和结构上的巨大差异，需要为每个不同的基因组开发一组特征明确的基因和剪接信号的训练集。此外，通常用于增强从头算基因预测的剪接比对的适用性受到限制，这是因为与已知基因的序列不具有相似性的基因的比例很高。大约一半推定的Pc基因没有确定的同源基因(这种情况类似于其他真菌，如酵母和脉孢子菌)。因此，在Pc和其他真菌基因组中寻找和注释基因是一个重大的挑战。在本提案中，作为通向完整注释的第一步，将开发软件和分析工具来识别Pc基因组中的假定基因。我们将利用EST数据库、可用的基因组序列、已知的Pc基因，以及这一提案中人员的专业知识，为Pc中的基因发现和注释创建一个集成的生物计算策略。我们的具体目标是：1)建立一个具有代表性的Pc基因数据库，将识别内含子/外显子边界和其他相关信号；2)开发和训练Pc特有的基因识别方法，采用分层策略，以一种新颖的方式结合先进的模式识别方法，如支持向量机、隐马尔可夫模型和自适应神经网络。

DESCRIPTION (provided by applicant): A Pneumocystis Genome Project was funded in 1999 with the goals of creating a physical map, an EST database and the assembled DNA sequence for the fungal pathogen, Pneumocystis carinii (Pc). Pc causes one of the major HIV-associated infections, Pneumocystis pneumonia (PcP). The genomic sequencing of Pc is ongoing: large contiguous stretches of several chromosomes as well as a non-redundant set of -1900 EST's are available for analysis. Strategies for gene finding in Pc and further annotation of gene products need to be developed now to fully utilize the growing body of sequence data. In this complementary project, we propose (in accord with specific goals of PAS-02-046) to develop and apply computational tools for gene prediction and annotation for the Pc genome in order to facilitate the search for novel drug targets and potential drug candidates for PcP. Despite the growing number of fungal genomes that are being sequenced, only few gene prediction programs for fungi have been developed. Due to specific biases in A/T content, intron/exon boundaries, promoter sequences and gene densities as well as large differences in organization and structure between different fungal genomes, a training set of well characterized genes and splicing signals needs to be developed for each distinct genome. Moreover, the applicability of splicing alignments that are commonly used to enhance ab initio gene predictions is limited due to the high percentage of genes that do not share similarity with sequences of known genes. Approximately half of the putative Pc genes have no identified orthologs (a situation similar to other fungi, such as yeast and Neurospora). Therefore, gene finding and annotation in Pc as well as in other fungal genomes represent a significant challenge. In the present proposal, as a first step toward full annotation, software and analysis tools will be developed to identify putative genes in the Pc genome. We will take advantage of the EST database, available genomic sequences, known Pc genes, as well as the expertise of the personnel on this proposal to create an integrated biological-computational strategy for gene finding and annotation in Pc. Our specific aims are: 1) To build a representative Pc gene database that will identify intron/exon boundaries and other relevant signals; 2) To develop and train Pc-specific gene recognition methods using hierarchical strategy that combines in a novel way advanced pattern recognition approaches such as Support Vector Machines, Hidden Markov Models and adaptable Neural Networks.