Computational exploration of cis-regulatory modules in rhythmic expression data using the “Exploration of Distinctive CREs and CRMs” (EDCC) and “CRM Network Generator” (CNG) programs

Understanding the effect of cis-regulatory elements (CRE) and clusters of CREs, which are called cis-regulatory modules (CRM), in eukaryotic gene expression is a challenge of computational biology. We developed two programs that allow simple, fast and reliable analysis of candidate CREs and CRMs that may affect specific gene expression and that determine positional features between individual CREs within a CRM. The first program, “Exploration of Distinctive CREs and CRMs” (EDCC), correlates candidate CREs and CRMs with specific gene expression patterns. For pairs of CREs, EDCC also determines positional preferences of the single CREs in relation to each other and to the transcriptional start site. The second program, “CRM Network Generator” (CNG), prioritizes these positional preferences using a neural network and thus allows unbiased rating of the positional preferences that were determined by EDCC. We tested these programs with data from a microarray study of circadian gene expression in Arabidopsis thaliana. Analyzing more than 1.5 million pairwise CRE combinations, we found 22 candidate combinations, of which several contained known clock promoter elements together with elements that had not been identified as relevant to circadian gene expression before. CNG analysis further identified positional preferences of these CRE pairs, hinting at positional information that may be relevant for circadian gene expression. Future wet lab experiments will have to determine which of these combinations confer daytime specific circadian gene expression.解析顺式调控元件（cis-regulatory elements, CRE）及其簇（即顺式调控模块cis-regulatory modules, CRM）对真核基因表达的调控效应，是计算生物学领域的一项挑战。我们开发了两款程序，可对可能影响特定基因表达、且可解析单个CRE在CRM内位置特征的候选CRE与CRM进行简便、快速且可靠的分析。第一款程序为「探索独特顺式调控元件与顺式调控模块」（"Exploration of Distinctive CREs and CRMs", EDCC），可将候选CRE与CRM与特定基因表达模式进行关联分析；针对CRE对，该程序还可确定单个CRE之间以及相对于转录起始位点的位置偏好性。第二款程序为「顺式调控模块网络生成器」（"CRM Network Generator", CNG），可通过神经网络对上述位置偏好性进行优先级排序，从而对EDCC所确定的位置偏好性实现无偏评估。我们利用拟南芥（Arabidopsis thaliana）昼夜节律基因表达的微阵列研究数据对这两款程序进行了验证。通过分析超过150万对CRE组合，我们筛选得到22组候选组合，其中多组包含已知的生物钟启动子元件，以及此前未被证实与昼夜节律基因表达相关的调控元件。CNG分析进一步明确了这些CRE对的位置偏好性，暗示存在可能与昼夜节律基因表达相关的位置调控信息。后续仍需通过湿实验验证，以确定其中哪些组合可赋予基因日间特异性的昼夜节律表达模式。