Huerta Araceli M, Collado-Vides Julio
Program of Computational Genomics, Nitrogen Fixation Center, UNAM, Cuernavaca, AP 565-A, Morelos 62100, Mexico.
J Mol Biol. 2003 Oct 17;333(2):261-78. doi: 10.1016/j.jmb.2003.07.017.
We present here a computational analysis showing that sigma70 house-keeping promoters are located within zones with high densities of promoter-like signals in Escherichia coli, and we introduce strategies that allow for the correct computer prediction of sigma70 promoters. Based on 599 experimentally verified promoters of E.coli K-12, we generated and evaluated more than 200 weight matrices optimizing different criteria to obtain the best recognition matrices. The alignments generating the best statistical models did not fully correspond with the canonical sigma70 model. However, matrices that correspond to such a canonical model performed better as tools for prediction. We tested the predictive capacity of these matrices on 250 bp long regions upstream of gene starts, where 90% of the known promoters occur. The computational matrix models generated an average of 38 promoter-like signals within each 250 bp region. In more than 50% of the cases, the true promoter does not have the best score within the region. We observed, in fact, that real promoters occur mostly within regions with high densities of overlapping putative promoters. We evaluated several strategies to identify promoters. The best one uses an intrinsic score of the -10 and -35 hexamers that form the promoter as well as an extrinsic score that uses the distribution of promoters from the start of the gene. We were able to identify 86% true promoters correctly, generating an average of 4.7 putative promoters per region as output, of which 3.7, on average, exist in clusters, as a series of overlapping potentially competing RNA polymerase-binding sites. As far as we know, this is the highest predictive capability reported so far. This high signal density is found mainly within regions upstream of genes, contrasting with coding regions and regions located between convergently transcribed genes. These results are consistent with experimental evidence that show the existence of multiple overlapping promoter sites that become functional under particular conditions. This density is probably the consequence of a rich number of vestiges of promoters in evolution. We suggest that transcriptional regulators as well as other functional promoters play an important role in keeping these latent signals suppressed.
我们在此展示了一项计算分析，结果表明σ70管家启动子位于大肠杆菌中具有高密度启动子样信号的区域内，并且我们介绍了能够正确地通过计算机预测σ70启动子的策略。基于大肠杆菌K-12的599个经实验验证的启动子，我们生成并评估了200多个权重矩阵，这些矩阵针对不同标准进行了优化，以获得最佳识别矩阵。生成最佳统计模型的比对并不完全符合经典的σ70模型。然而，与这种经典模型相对应的矩阵作为预测工具表现得更好。我们在基因起始上游250 bp长的区域测试了这些矩阵的预测能力，已知90%的启动子位于该区域。计算矩阵模型在每个250 bp区域内平均产生38个启动子样信号。在超过50%的情况下，真正的启动子在该区域内并非得分最高。事实上，我们观察到真正的启动子大多出现在具有高密度重叠假定启动子的区域内。我们评估了几种识别启动子的策略。最佳策略使用构成启动子的-10和-35六聚体的内在得分以及利用从基因起始处开始的启动子分布的外在得分。我们能够正确识别86%的真正启动子，每个区域平均产生4.7个假定启动子作为输出，其中平均有3.7个以簇的形式存在，作为一系列重叠的潜在竞争性RNA聚合酶结合位点。据我们所知，这是迄今为止报道的最高预测能力。这种高信号密度主要在基因上游区域被发现，这与编码区域以及位于反向转录基因之间的区域形成对比。这些结果与实验证据一致，实验证据表明存在多个重叠的启动子位点，这些位点在特定条件下会发挥功能。这种密度可能是进化过程中大量启动子遗迹的结果。我们认为转录调节因子以及其他功能性启动子在抑制这些潜在信号方面发挥着重要作用。
