Behkamal Bahareh, Naghibzadeh Mahmoud, Pagnani Andrea, Saberi Mohammad Reza, Al Nasr Kamal
Department of Computer Engineering, Faculty of Engineering, Ferdowsi University of Mashhad, Mashhad, 9177948944, Iran.
Department of Applied Science and Technology (DISAT), Politecnico di Torino, Corso Duca degli Abruzzi 24, Torino, Italy; Italian Institute for Genomic Medicine (IIGM), IRCC-Candiolo, Candiolo, TO, Italy; INFN Sezione di Torino, Via P. Giuria 1, Torino, Italy.
J Mol Graph Model. 2021 Mar;103:107815. doi: 10.1016/j.jmgm.2020.107815. Epub 2020 Nov 28.
Cryo-electron microscopy (cryo-EM) has recently emerged as a prominent biophysical method for macromolecular structure determination. Many research efforts have been devoted to produce cryo-EM images, density maps, at near-atomic resolution. Despite many advances in technology, the resolution of the generated density maps may not be sufficiently adequate and informative to directly construct the atomic structure of proteins. At medium-resolution (∼4-10 Å), secondary structure elements (α-helices and β-sheets) are discernible, whereas finding the correspondence of secondary structure elements detected in the density map with those on the sequence remains a challenging problem. In this paper, an automatic framework is proposed to solve α-helix correspondence problem in three-dimensional space. Through modeling of the sequence with the aid of a novel strategy, the α-helix correspondence problem is initially transformed into a complete weighted bipartite graph matching problem. An innovative correlation-based scoring function based on a well-known and robust statistical method is proposed for weighting the graph. Moreover, two local optimization algorithms, which are Greedy and Improved Greedy algorithms, have been presented to find α-helix correspondence. A widely used data set including 16 reconstructed and 4 experimental cryo-EM maps were chosen to verify the accuracy and reliability of the proposed automatic method. The experimental results demonstrate that the automatic method is highly efficient (86.25% accuracy), robust (11.3% error rate), fast (∼1.4 s), and works independently from cryo-EM skeleton.
低温电子显微镜（cryo-EM）最近已成为一种用于确定大分子结构的重要生物物理方法。许多研究致力于生成近原子分辨率的低温电子显微镜图像、密度图。尽管技术上有许多进步，但生成的密度图的分辨率可能仍不足以直接构建蛋白质的原子结构，也缺乏足够的信息。在中等分辨率（约4 - 10埃）下，可以辨别二级结构元件（α螺旋和β折叠），然而，在密度图中检测到的二级结构元件与序列上的二级结构元件之间找到对应关系仍然是一个具有挑战性的问题。本文提出了一个自动框架来解决三维空间中的α螺旋对应问题。通过借助一种新颖的策略对序列进行建模，α螺旋对应问题最初被转化为一个完全加权二分图匹配问题。提出了一种基于一种著名且稳健的统计方法的创新型基于相关性的评分函数来对图进行加权。此外，还提出了两种局部优化算法，即贪心算法和改进贪心算法，以找到α螺旋对应关系。选择了一个广泛使用的数据集，包括16个重建的和4个实验性低温电子显微镜图，以验证所提出的自动方法的准确性和可靠性。实验结果表明，该自动方法高效（准确率达86.25%）、稳健（错误率为11.3%）、快速（约1.4秒），并且独立于低温电子显微镜骨架工作。
