Improved accuracy assessment for 3D genome reconstructions
Improved accuracy assessment for 3D genome reconstructions
复制标题
DOI：
10.1186/s12859-018-2214-2
复制
发表时间：
2018-05-30
影响因子：
3
通讯作者：
Bengtsson, Henrik L.
中科院分区：
文献类型：
--
作者：
Segal, Mark R.;Bengtsson, Henrik L.
背景：三维（3D）基因组空间组织对于包括转录在内的许多细胞功能至关重要，而某些构象驱动的结构改变通常是致癌的。基因组构象一直难以阐明，但染色质构象捕获测定的出现，特别是Hi-C，已经改变了对染色质结构的理解，并产生了许多生物学见解。尽管这些发现大多来自于对这些检测产生的接近数据的分析，但产生三维重建的附加价值已经被证明，部分来自于将基因组特征叠加在重建上。然而，基于三维结构的分析的优势显然取决于随之而来的重建的准确性，这很难评估。竞争性重建算法的支持者通过模拟玩具结构和/或有限的荧光原位杂交（FISH）成像来评估其准确性，该成像具有少量低分辨率探针。因此，需要新的重建精度评估方法。结果：在这里，我们利用最近设计的两种分析方法来开发评估3D重建精度的方法。多重FISH将探针数量增加了一个数量级，因此探针间距离的数量增加了两个数量级，通过均方偏差（MSD）为结构级评估提供了足够的信息。至关重要的是，强调多重FISH应用程序的是大量与坐标系对齐的复制，这些复制为MSD统计数据的参考分布提供了基础。使用该系统，我们发现基于Hi-C数据的IMR90细胞重建对某些染色体是准确的，而对其他染色体则不准确。第二种新的检测方法，基因组结构定位，利用大量的薄冷冻切片来获得接近度的测量。我们利用冰冻切片的平面度-不用于推断接近度-获得重建精度的措施，通过重采样提供的参考。应用于小鼠胚胎干细胞显示重建准确性随染色体而异。结论：我们已经开发了评估3D基因组重建准确性的方法，这些方法利用了最近先进的多重FISH和基因组结构定位分析的特征。这些方法可以帮助克服缺乏黄金标准来进行此类评估，鉴于围绕3D基因组重建的相当大的不确定性，这一点很重要。
Background: Three dimensional (3D) genome spatial organization is critical for numerous cellular functions, including transcription, while certain conformation-driven structural alterations are frequently oncogenic. Genome conformation had been difficult to elucidate but the advent chromatin conformation capture assays, notably Hi-C, has transformed understanding of chromatin architecture and yielded numerous biological insights. Although most of these findings have flowed from analysis of proximity data produced by these assays, added value in generating 3D reconstructions has been demonstrated, deriving, in part, from superposing genomic features on the reconstruction. However, advantages of 3D structure-based analyses are clearly conditional on the accuracy of the attendant reconstructions, which is difficult to assess. Proponents of competing reconstruction algorithms have evaluated their accuracy by recourse to simulation of toy structures and/or limited fluorescence in situ hybridization (FISH) imaging that features a handful of low resolution probes. Accordingly, new methods of reconstruction accuracy assessment are needed.Results: Here we utilize two recently devised assays to develop methodology for assessing 3D reconstruction accuracy. Multiplex FISH increases the number of probes by an order of magnitude and hence the number of inter-probe distances by two orders, providing sufficient information for structure-level evaluation via mean-squared deviations (MSD). Crucially, underscoring multiplex FISH applications are large numbers of coordinate-system aligned replicates that provide the basis for a referent distribution for MSD statistics. Using this system we show that reconstructions based on Hi-C data for IMR90 cells are accurate for some chromosomes but not others. The second new assay, genome architecture mapping, utilizes large numbers of thin cryosections to obtain a measure of proximity. We exploit the planarity of the cryosections - not used in inferring proximity - to obtain measures of reconstruction accuracy, with referents provided via resampling. Application to mouse embryonic stem cells shows reconstruction accuracies that vary by chromosome.Conclusions: We have developed methods for assessing the accuracy of 3D genome reconstructions that exploit features of recently advanced multiplex FISH and genome architecture mapping assays. These approaches can help overcome the absence of gold standards for making such assessments which are important in view of the considerable uncertainties surrounding 3D genome reconstruction.