New methods and enhanced software for predicting functional SNPs

描述（由申请人提供）：DNA测序技术的快速发展使研究和临床环境中人类等位基因变异的大量鉴定和编目成为可能。当今人类遗传学面临的一个关键挑战是在无数等位基因中识别那些对分子功能和表型有影响的变异。我们早期开发了预测人类突变和非同义snp功能效应的计算方法，并在PolyPhen和随后的PolyPhen-2软件工具中实现了这些方法。我们在实验室中维护这些计算工具的在线和独立版本。这些工具被遗传学家广泛用于各种研究和临床应用。大规模人口测序项目的爆发极大地增加了对预测方法的需求。这些项目也对方法的重大改进和裁剪软件以适应新技术先进的人类遗传学的具体应用提出了新的要求。具体来说，大规模的外显子组测序项目旨在鉴定包含与人类表型相关的罕见编码变异的基因，需要高度准确、易于使用和快速的方法来注释大量序列变异。另一方面，DNA测序正迅速成为临床遗传诊断的一种选择方法。人类疾病基因新序列变异的解释成为测序数据诊断分析的主要瓶颈。临床遗传诊断的应用需要大幅提高预测方法的准确性，并开发针对特定蛋白质组的方法，并产生针对个别诊断测试的预测。当前对序列变异解释的需求与大大增强计算方法和软件的机会并行。多种脊椎动物的基因组为预测提供了丰富的信息资源。需要新的统计方法来最佳地利用这些数据。最近人类突变和常见snp数据库规模的增加提供了更大的训练和测试数据集。应该开发新的方法，以充分受益于大量的训练和测试数据。在具体目标1中，我们将开发一种由系统发育树指导的预测方法，该方法将利用脊椎动物基因组的比对。我们将进一步在比较基因组学数据分析中纳入氨基酸位置之间的相互作用，以考虑代偿替代。在Specific Aim 2中，我们将开发一个版本的PolyPhen软件，用于分析外显子组或基因组测序数据集。我们将在统计测试中整合功能预测，以检测罕见的非同义变异的表型关联。在Specific Aim 3中，我们将与临床遗传学家密切合作，测试开发专门用于个体诊断测试的预测方法的可行性，这些方法将达到临床有用的特异性和敏感性水平。

DESCRIPTION (provided by applicant): Rapid advances in DNA sequencing technology enabled massive identification and cataloging of human allelic variation in research and clinical setting. A key challenge for human genetics today is to identify, among the myriad of alleles, those variants that have an effect on molecular function and phenotypes. We earlier developed computational methods for predicting the functional effect of human mutations and non-synonymous SNPs and implemented these methods in software tools PolyPhen and subsequently PolyPhen-2. We maintain both online and standalone versions of these computational tools in our laboratory. These tools are widely used by geneticists in a variety of research and clinical applications. Explosion of large-scale population sequencing projects greatly increased demand for the prediction methods. These projects also set new requirements for significant improvements of the methods and for tailoring software to specific applications in new technologically advanced human genetics. Specifically, massive exome sequencing projects aiming at identifying genes that harbor rare coding variants involved in human phenotypes require highly accurate, easy to use and fast methods for annotating large numbers of sequence variants. On the other hand, DNA sequencing is rapidly becoming a method of choice in clinical genetic diagnostics. Interpretation of novel sequence variants in human disease genes becomes the major bottleneck in diagnostic analysis of sequencing data. Applications to clinical genetic diagnostics require substantial increase in the accuracy of prediction methods and development of methods that target specific protein groups and generate predictions specific to individual diagnostic tests. The current need in interpretation of sequence variants is paralleled by the opportunity to greatly enhance computational methods and software. Genomes of multiple vertebrates provide a rich resource of information for generating predictions. New statistical approaches are needed to optimally employ these data. Recent increase of the size of databases of human mutations and common SNPs provide much larger training and testing datasets. New methods should be developed to fully benefit from large training and testing data. In Specific Aim 1, we will develop a prediction method guided by the phylogenetic tree that would utilize alignments of vertebrate genomes. We will further incorporate interactions between amino acid positions in the analysis of comparative genomics data to take into account compensatory substitutions. In Specific Aim 2, we will develop a version of PolyPhen software for the analysis of exome or genome sequencing datasets. We will integrate functional predictions in the statistical tests to detect phenotypic association of rare non- synonymous variants. In Specific Aim 3, in close collaboration with clinical geneticists we will test feasibility of developing prediction methods specialized for individual diagnostic tests that would achieve clinically useful levels of specificity and sensitivity.