当前位置: X-MOL 学术Acta Cryst. D › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Deep-learning map segmentation for protein X-ray crystallographic structure determination.
Acta Crystallographica Section D ( IF 2.6 ) Pub Date : 2024-06-27 , DOI: 10.1107/s2059798324005217
Pavol Skubák 1
Affiliation  

When solving a structure of a protein from single-wavelength anomalous diffraction X-ray data, the initial phases obtained by phasing from an anomalously scattering substructure usually need to be improved by an iterated electron-density modification. In this manuscript, the use of convolutional neural networks (CNNs) for segmentation of the initial experimental phasing electron-density maps is proposed. The results reported demonstrate that a CNN with U-net architecture, trained on several thousands of electron-density maps generated mainly using X-ray data from the Protein Data Bank in a supervised learning, can improve current density-modification methods.

中文翻译:


用于蛋白质 X 射线晶体结构测定的深度学习图分割。



当从单波长异常衍射 X 射线数据求解蛋白质结构时,通过从异常散射子结构定相获得的初始相位通常需要通过迭代电子密度修改来改进。在这份手稿中,提出了使用卷积神经网络(CNN)对初始实验定相电子密度图进行分割。报告的结果表明,具有 U-net 架构的 CNN 在监督学习中主要使用来自蛋白质数据库的 X 射线数据生成的数千个电子密度图上进行训练,可以改进当前的密度修改方法。
更新日期:2024-06-27
down
wechat
bug