New methods and enhanced software for predicting functional SNPs

 描述（由申请人提供）：基因组学研究的重点正在迅速从遗传变异数据的积累转移到等位基因变异的功能解释。测序研究正在成为遗传学所有领域的标准方法，对预测突变功能效应的计算方法产生了前所未有的需求。我们不断开发和维护PolyPhen-2，这是一种用于预测错义突变功能效应的计算方法。PolyPhen-2基于比较序列分析和蛋白质结构分析进行预测。这种方法被广泛应用于遗传学的各个领域。尽管用户基数很大，我们也在不断努力提高预测精度，但仍有很大的改进空间，而且非常需要提高该方法的精度。我们最近对有害等位基因群体遗传学的研究指出了有害变异分析和预测的基本复杂性。对这些复杂性的更好理解，新类型的训练和验证数据以及算法方法使我们能够大幅改进计算方法和软件。我们还将通过解决以前服务不足的需求来扩大该方法的效用。基因发现研究在基因和变体水平上优先考虑已鉴定的变体。PolyPhen-2和其他预测方法目前解决的问题是给定的变体是否可能影响基因功能。同样重要的考虑因素是， 是否含有这种变异是一种病态基因，以及这种基因或结构域中的大多数错义变化是否可能具有功能影响。深度群体测序数据以及已知疾病变体的目录可以与进化和结构分析一起使用，以优先考虑基因。许多大规模测序项目正在从外显子组向全基因组测序过渡。这为非编码变异的分析打开了一个视角。非编码变异在多基因复杂表型的遗传学中起着关键作用。然而，对于以孟德尔方式分离的表型，大效应非编码变体的重要性尚不清楚，并且仍在争论中。通过对与已知基因座相关但缺乏蛋白质编码变体的孟德尔疾病病例进行单独支持的全基因组测序，我们将以无偏的方式选择非编码孟德尔突变，分析潜在的潜在生物学，并开发计算预测因子。这种方法是从根本上不同于现有的非编码变异的分析，预测保守或减少序列多样性，而不是直接确定的致病作用的努力。在具体目标1中，我们将对预测突变功能效应的计算方法进行实质性改进，并将这些改进纳入PolyPhen软件。在具体目标2中，我们将根据人口和疾病遗传学数据开发基于基因的评分，并将其与基于变异的预测相结合。在具体目标3中，我们将预测扩展到非编码变化。

 DESCRIPTION (provided by applicant): The focus of genomics research is rapidly shifting from the accumulation of genetic variation data to the functional interpretation of allelic variant. Sequencing studies are becoming the standard approach in all areas of genetics, generating an unprecedented demand for computational methods to predict the functional effect of mutations. We continuously develop and maintain PolyPhen-2, a computational method for predicting the functional effect of missense mutations. PolyPhen-2 makes predictions based on comparative sequence analysis and analysis of protein structure. This method is being widely applied in diverse areas of genetics. In spite of the large user base and our continuing efforts to increase prediction accuracy, there is an ample room for improvement and a great need to improve accuracy of the method. Our recent studies on population genetics of deleterious alleles point to fundamental complexities in the analysis and prediction of deleterious variation. The improved understanding of these complexities, new types of training and validation data and algorithmic approaches position us to substantially improve the computational method and the software. We will also expand the utility of the method by addressing previously underserved needs. Gene discovery studies prioritize identified variants both at the gene and the variant level. The question currently addressed by PolyPhen-2 and other prediction methods is whether a given variant is likely to affect gene function. Equally important considerations are whether a gene that harbors this variant is a morbid gene and whether most missense changes in this gene or a domain are likely to have a functional impact. Deep population sequencing data together with catalogs of known disease variants can be used in concert with evolutionary and structural analyses to prioritize genes. Many large-scale sequencing projects are transitioning from exomes to whole genome sequencing. This opens a perspective for the analysis of non-coding variation. Non-coding variation has been shown to play a key role in genetics of polygenic complex phenotypes. However, the importance of large effect non-coding variants for phenotypes that segregate in the Mendelian fashion is unclear and still under debate. Through separately supported whole genome sequencing of cases of Mendelian diseases linked to known loci but lacking protein-coding variants we will select non-coding Mendelian mutations in an unbiased fashion, analyze the potential underlying biology and will develop a computational predictor. This approach is fundamentally different from existing efforts on the analysis of non-coding variation that predict conservation or a reduction in sequence diversity rather than directly ascertain the pathogenic effect. In Specific Aim 1 we will make substantial improvements in computational methods for predicting the functional effect of mutations and incorporate these improvements into the PolyPhen software. In Specific Aim 2 we will develop gene-based scores based on population and disease genetics data and integrate them with the variant-based predictions. In Specific Aim 3 we will extend the prediction to non-coding variation.