Inferring gene regulatory circuitry from functional genomics data

描述(由申请人提供)： 项目概述很早就知道基因组DNA的甲基化影响基因表达。然而，潜在的结构性机制在很大程度上仍然模糊不清。在这个项目中，我们将寻求一种新的策略来预测甲基化如何影响转录因子(TF)结合，从而影响局部染色质结构和每个细胞特征的基因表达的复杂的全基因组格局。我们将探索这样的假设，即甲基化导致DNA形状的局部变化，这反过来又改变了TF结合亲和力。动机来自我们最近对核酸内切酶I内在特异性的分析。我们发现，胞嘧啶甲基化极大地提高了DNase I裂解邻近CpG二核苷酸的DNA骨架的速度。对此的解释是，在主槽中增加一个甲基导致DNA形状的变化，局部缩小了小槽，并增强了DNA的负主链磷酸盐与DNA酶I的正氨基酸残基之间的静电相互作用。通过小槽识别DNA形状也有助于真核生物TF的结合特异性，这表明可以通过对未甲基化DNA序列中TF结合偏好的形状分析来预测甲基化敏感性，这是可以获得充足的高通量体外结合数据的。为了探索这一点，我们将首先开发和拟合整合DNA碱基和形状读出的TF结合特异性模型，方法是扩展FeatureREDUCE算法背后的生物物理模型，以包括来自自由DNA分子的计算机模拟的DNA形状信息。接下来，我们将使用这些整合的碱基/形状识别模型来预测TF的甲基化敏感性，并通过实验验证这些预测。在一种平行的方法中，我们将扩展我们最近开发的SELEX-SEQ方法，使用甲基化和未甲基化的DNA配体的条形码混合物来创建甲基化对具有代表性的一组TF结合亲和力的影响的详细地图。最后，我们将分析 利用家族水平的建模，Tf的结合专一性如何取决于其氨基酸序列。使用来自同一结构Tf家族的大量TF的生物物理碱基和形状识别参数，以及碱基偏好的新的几何表示，我们将预测当氨基酸残基发生突变时，碱性螺旋-环-螺旋(BHLH)和碱性亮氨酸拉链(BZIP)蛋白质的结合特异性如何变化，并在实验上验证这些预测。我们将使用相同的基于家族的方法来证明bHLH因子存在替代的二聚体结合模式，并研究是否可以从其蛋白质序列预测TF使用这些替代模式的倾向。

DESCRIPTION (provided by applicant): PROJECT SUMMARY It has long been known that methylation of genomic DNA influences gene expression. The underlying structural mechanisms, however, largely remain obscure. In this project, we will pursue a new strategy for predicting how methylation affects transcription factor (TF) binding, thereby influencing the intricate genomewide landscape of local chromatin structure and gene expression that characterizes each cell. We will explore the hypothesis that methylation causes local changes in DNA shape, which in turn modify TF binding affinity. Motivation comes from our recent analysis of the intrinsic specificity of the endonuclease DNase I. We found that cytosine methylation greatly increases the rate at which DNase I cleaves the DNA backbone adjacent to CpG dinucleotides. The explanation for this is that adding a methyl group in the major groove causes changes in DNA shape that locally narrow the minor groove and enhance the electrostatic interaction between negative backbone phosphates of the DNA and positive amino-acid residues of DNase I. Recognition of DNA shape via the minor groove can also contribute to the binding specificity of eukaryotic TFs, suggesting that methylation sensitivity can be predicted from a shape-based analysis of TF binding preferences among unmethylated DNA sequences, for which ample high-throughput in vitro binding data is available. To explore this, we will first develop and fit models of TF binding specificity that integrate DNA base and shape readout by extending the biophysical model underlying our FeatureREDUCE algorithm to include information about DNA shape from computer simulations of free DNA molecules. Next, we will use these integrated base/shape recognition models to make predictions regarding the methylation sensitivity of TFs, and validate these experimentally. In a parallel approach, we will extend our recently developed SELEX-seq method by using barcoded mixtures of methylated and unmethylated DNA ligands to create detailed maps of the effect of methylation on binding affinity for a representative set of TFs. Finally, we will analyze how the binding specificity of a TF depends on its amino-acid sequence using family-level modeling. Using biophysical base and shape recognition parameters estimated for a large number of TFs from the same structural TF family, along with a novel geometric representation of base preference, we will predict how the binding specificity of basic helix-loop-helix (bHLH) and basic leucine zipper (bZIP) proteins changes when amino-acid residues are mutated, and experimentally validate these predictions. We will use the same family-based approach to demonstrate the existence of alternative dimeric binding modes for bHLH factors, and investigate whether the propensity of a TF to use these alternative modes can be predicted from its protein sequence.