DNA PATTERN IDENTIFICATION AND ANALYSIS

描述（由申请人提供）： 转录调控是所有细胞正常运作的核心。确定所有转录因子（TF）的NA结合位点将极大地促进我们对伴随遗传差异的正常和疾病状态下基因表达的调控网络和变异的理解。新的高通量技术正在以大大增加的速率生成关于转录因子的DNA结合特异性的数据，但是需要良好的计算方法来最大化从这些数据中提取的生物信息。在上一个资助期，我们开发了新的和改进的方法，用于分析三种不同类型的高通量特异性数据。在本提案中，我们将以几种方式扩展这些方法，包括分析其他类型数据的方法，以及开发更复杂的模型，这些模型是充分表示某些因素的特异性所必需的。对于其特异性不能很好地由位置权重矩阵（PWM）表示的TF，需要更复杂的模型，所述位置权重矩阵（PWM）施加约束，即结合位点内的位置独立地贡献于结合。我们将为TF开发模型，这些TF允许更高阶的相互作用，以及可以以替代模式结合并需要多个独立模型来表示它们的TF。将改进的模型与TF的体内定位分析进行比较，以更好地评估哪些结合位点是间接的或需要与其他因子协同结合。我们还利用大量增加的数据来开发改进的识别模型，该模型可以基于蛋白质序列预测TF的特异性，并有助于设计具有新特异性的新因子。这将首先对同源结构域和锌指蛋白进行，这是真核基因组中两个最大的TF家族，并且具有最可用的特异性信息。我们还将利用细菌基因组的大量信息来开发各种细菌TF家族的特异性模型。一种新的实验方法将被用来更全面地评估蛋白质残基和结合位点碱基对之间的非独立相互作用，这将导致识别建模的进一步改进。我们继续与实验生物学家合作，这有助于他们使用我们的计划并进一步实现他们的研究目标，并帮助我们确定当前方法的局限性并促进改进。我们还进行了一项新的合作，旨在改进基于分子建模预测蛋白质-DNA相互作用特异性的方法，将他们在热力学和结构建模方面的专业知识与我们广泛的TF结合特异性模型相结合。

DESCRIPTION (provided by applicant): The regulation of transcription is central to the proper functioning of all cells. Identifying the NA binding sites for all transcription factors (TFs) would greatly facilitate our understanding of regulatory networks and variations in gene expression, both normal and in disease states, that accompany genetic differences. New high-throughput technologies are generating data about the DNA binding specificity of transcription factors at a greatly increased rate, but good computational methods are required to maximize the biological information extracted from those data. In the previous funding period we developed new, and improved, methods for the analysis of three different types of high- throughput specificity data. In this proposal we will expand on those methods in several ways, including methods for analyzing additional types of data and the development of more complex models that are required for the adequate representation of the specificity of some factors. More complex models are needed for TFs whose specificity is not well represented by position weight matrices (PWMs) which impose the constraint that the positions within the binding site contribute independently to the binding. We will develop models for TFs that allow for higher-order interactions as well as for TFs that can bind in alternative modes and require multiple, independent models to represent them. The improved models will be compared to in vivo location analysis for TFs to better assess which binding sites are indirect or require cooperative binding with other factors. We also take advantage of greatly increased data to develop improved recognition models that can predict the specificity of TFs based on the protein sequence and aid in the design of new factors with novel specificity. This will be done initially for homeodomain and zinc finger proteins, the two largest families of TFs in eukaryotic genomes and the ones with the most available specificity information. We will also take advantage of the vast information available for bacterial genomes to develop specificity models for various bacterial TF families. A new experimental method will be employed to more comprehensively assess the non-independent interactions between protein residues and binding site base-pairs, which should lead to further improvements in recognition modeling. We continue collaborating with experimental biologists, which helps them use our programs and further their research goals, and helps us identify the limitations of the current methods and fosters improvements. We also have a new collaboration that seeks to improve upon methods for predicting specificity in protein-DNA interactions based on molecular modeling, combining their expertise in thermodynamic and structural modeling with our extensive models of TF binding specificity.