%0 Journal Article
%T Deep-learning map segmentation for protein X-ray crystallographic structure determination.
%A Skubák P
%J Acta Crystallogr D Struct Biol
%V 80
%N 0
%D 2024 Jul 1
%M 38935341
%F 5.699
%R 10.1107/S2059798324005217
%X 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.